LEARNING PROCESSES

The previous chapters have all dealt with information technology and the emergence of a learning technology that is both powerful and responsive, in other words a useful technology for learning. These chapters necessarily dealt with the phenomenon of learning, but always indirectly or implicitly. Now is the time to explore human learning explicitly, as a backdrop for learning technology. This therefore is a chapter dealing with psychology, and more specifically with the cognitive processes that constitute learning.

THE NEED FOR AN ANALYTICAL VIEW

A vexing problem in dealing with learning is the inadequacy of our vocabulary to describe the phenomena associated with the notion of learning. The term 'learning' is too broad: learning how to ski and learning trigonometry are both 'learning', but yet we feel they are rather different phenomena; likewise for learning not to touch hot light bulbs and learning the song `O sole mio'. The products of learning are also confusing: we learn knowledge and we learn skills, but skills often include or involve knowledge. And so on with many terms related to learning. And then of course, there are also the broad theoretical perspectives that shape how we talk about learning: whether we treat it as a change in behavior, as the behaviorists did in the mid-century, or as a cognitive change instead, as psychology has tended to do since then.

This confusion is related to the question of whether we need to focus on many different forms of learning, or instead on only the core processes of learning that operate in an underlying manner irrespective of form of learning. And of course, how we describe learning is all quite arbitrary and relative: at a given level of analysis, we can emphasize differences between application areas, while still maintaining at another level that all learning shares basic core elements.

The greatest problem with the term 'learning', however, is that it is used to talk about both the internal and the external manifestations of learning. We use learning to refer to the occurrence of change itself (whether in behavior or in cognition does not matter here): "before, I couldn't do that (or didn't know that), and now I can do it (or I do know it)". Something happened internally that led to a learning result. But we also use learning in a loose way to refer to the process which brings about the change: repeating a poem over and over for instance, or connecting two ideas together in one's mind through imagery, and so on. This is what we engage in when we learn, the external manifestation of learning. This latter aspect of the notion of learning overlaps oftentimes with the notion of 'studying', as in the phrase 'learning chemistry' or 'learning how to sail'.

It is hard to avoid the potential confusions which are apt to crop up therefore in discussing learning. And yet, there is a core meaning to the term taken broadly which can hopefully be identified and characterized in useful terms. What is aimed for in the end is a global, coherent, and plausible description of the phenomenon as it might apply in different circumstances, that is, a theory of learning.

This chapter then presents a theory of learning, one which I have sometimes called a 'levels' theory of learning, for reasons which will become clear as we go on. In line with the rest of this book, there is no attempt here to base the theory on a foundation of empirical results from research in psychology. I am more interested in fitting the ideas together than in establishing the validity of the theory.

LEVELS OF FUNCTIONING

In the early part of this book, I discussed experiential learning and artificial realities, and in order to do so , I had to dimensionalize experience, from the sensual at one end, on to the analogue, and on yet to the symbolic at the other end (alternate terms used to characterize this dimension were enactive, iconic, and abstract). What the dimension represents is how close, or how remote, one is to the physical world in apprehending it, how 'touchable' the world is.

It is on this dimension that behaviorist and cognitive views of learning distance themselves from one another. The behaviorist's definition of learning as a change in behavior relates to the sensual/enactive level of functioning, whereas the cognitivist's view of learning as the elaboration of cognitive structures relates to the iconic and to the symbolic levels of functioning. Also of interest is the positioning on this dimension of the three forms of learning which are commonly distinguished: motor and affective learning relate to the sensual side of the dimension, whereas conceptual learning, as its name directly implies, relates to the other side of the dimension.

Functioning in the world thus involves operating in more or less abstract terms depending on circumstances. It involves acting concretely with the hard stuff of this world, or acting imaginatively with the soft stuff of the world. One goes from hard sensation and behavior to internal representation of this hard stuff, first in analogue terms, and then in symbolic terms. These soft elements can then be elaborated upon and combined through thought into more and more complex structures portraying either real aspects of the world or slightly alternate versions of reality.

This progression along the sensual-symbolic dimension reflects the course of our evolutionary development, just as it also reflects the course of human development from infancy to adulthood, as described by Piaget who characterized it in a series of stages going from the sensory-motor stage to the final stage of formal reasoning. In each and every human, the progression is also present in our everyday dealings: we learn by feeling the world, by seeing images of it, and by thinking about it. We interact at different levels with the world. How does that impinge on learning? What cuts across those levels of functioning, if anything? That is our basic theoretical question. The answer lies in the notions of association and structure.

ASSOCIATION AND STRUCTURE

Association and structure are the cornerstones of learning, and the constructs that effectively cut across levels of human functioning. This is not surprising really, since behaviorism was built on associationism, and cognitivism largely emphasizes structure, the philosophical counterparts of each tradition going back to the dispute over knowledge between the empiricists and the rationalists. What may be less evident, however, is how association and structure relate to one another. That, they do through an underlying dimension based on coherence. As will be seen, this is the crux of the matter so far as learning is concerned.

In a nutshell, here is how learning occurs in mature adults. Being formal reasoners, when we are faced with novel information, we go for comprehending it; if that fails, or more precisely, in as much as that fails, then we go for rehearsing it. Thus, comprehension and rehearsal are the two processes through which we intentionally learn. Comprehension is a structural process, and rehearsal rests on an associative one. That we opt for comprehension over rehearsal whenever we can is natural since it is less demanding in cognitive terms. Which process occurs, however, is a function of the coherence involved in the learning situation.

This nutshell description of learning now needs to be elaborated. To do so, we will have to consider various notions surrounding the concept of learning, and in particular the concepts of memory and thought. But, as will be evident, association and structure will remain the cornerstones of the phenomenon of learning.

The result of learning is a changed memory. It is a mind that is slightly different than it was before learning occurred. Now, an interesting aspect of considering the whole mind as the human memory bank is that we must then conclude that learning occurs all the time that we interact with our world around us. So does forgetting, which keeps the mind lean and functionally ready for intellectual action.

When my wife tells me that we have been invited to a party, I have learned something: I have recorded that a part of my schedule is now fixed, that we will be seeing so and so, and so on. All this is learning, in the sense that my memory structure has changed. Fortunately, some time after the party, it will be forgotten. We often restrict the term learning to the recording in memory of information that will be potentially useful at many later points in life. This would include the traditional subject matters of schooling, as well as much other learning like riding a bicycle, using a word processor, and so on. We artificially restrict the term to important matters of long-term duration. The mind, however, makes no such distinction. It simply records experiences, irrespective of whether they are thought of highly or not. Learning is simply a recording process that leads to changes in one's memory bank. What we learn, however, and for how long, is another matter.

Memory is essentially an associative affair: it is a mass of associations which are stored away in our brains, the result of our experiences in the world (in effect, the result of all of our learning, just to phrase it in a circular manner), minus of course what has been forgotten. These associations are strengthened connections between elements of knowledge. The strength of the connection is of course variable: it depends directly dependent on the circumstances of learning, which will be discussed shortly.

Now, what is implied by all this is that, in a very real sense, memory is a jumble of associations reflecting our individual experience in the world. For instance, even though my knowledge of physics may appear to be more than just a jumble of associations, what about the fact that whenever I think of physics, I think of death (this physics - death association was laid in my mind by the death of my physics instructor). To speak of a jumble of associations makes memory appear as a random set of associations, which of course it is not. The associations were laid down (i.e.. learned) according to some rather specific rules involving structural features of our experience. However, that in no way confounds the view that our mind is no more than connections between ideas, i.e.. associations. Tons of them, admittedly, but only that.

The elements of knowledge that are thus connected in associations are varied. Some examples will illustrate this diversity:

Einstein - E=MC2

Vancouver - large city in western Canada

ANOVA - a procedure for a statistical test

Communism - the form of government in Russia

Revolution- an American historical experience

Russian revolution - 1917

What is seen here is that the units of knowledge are relative, both in scope and in perspective. In terms of scope, associations can exist between say a scent and a particular herb, or between cooking chicken and using herbs in general, without reference to any particular herb. In terms of perspective, one can associate a scent with the name of an herb ("that's oregano") or one can associate it with the visual shape of the bush. These are really different associations altogether, even if they each concern what is known as oregano, since they can be exclusive: one can exist without the other.

The relativity of knowledge is interesting, but it is not central to our understanding of learning. What is central, though, is the idea that, irrespective of what type of units of knowledge are involved in cognitive processing, the basic constituents of memory are associations of variable strength. Memory is associative! So what about structure?

Enter thought. Thought is essentially nothing more than a fancy word that refers to the process of modeling reality. Indeed, this is quite evident whenever one examines think-aloud protocols collected while people attempt some cognitive task: "now, let's see..., this is of this sort, so therefore..., but then, that would imply that..., so I think I'll try it this way...".

Modeling in turn is the process of building a structural representation of some aspect of reality, be it an object (such as a jet fighter), a process (the ecological balance of a park), or a procedure (how to send an electronic mail message). Modeling is sometimes mediated through iconic means (the plans of the jet fighter) or even through physical ones (a plastic model of the plane), but more often, it is simply symbolic (my idea of what a jet fighter constitutes). Symbolic modeling is pervasive: each one of our concepts is a symbolic model of some aspect of reality. Call them mental models, qualitative models, process models, schemata, cognitive structures, or whatever, they are basically symbolic models of aspects of reality.

Now here is the crux of the matter: we build these cognitive structures during learning and we employ them during remembering and problem-solving in everyday life, but they do not change the fact that memory is associative. Memory is not an internal structure that we inspect or manipulate, it is merely a set of connections between ideas. Our usage of the terms cognitive structures or mental models to refer to the set of associations related to a given domain (irrespective of its scope) is merely a manner of speech reflecting our prior experience in the world. Thus mental structures do not exist as such, only associations do. Whatever structure we 'see' in the mind lies merely in our manner of description.

What I have been describing here, it is important to note, is the result of learning, in effect the memory trace, and not learning itself, not the process of learning. It is in the latter that structure is generally so important, for it allows us to go beyond the associative level of cognitive functioning.

LEARNING

Learning can be described in a variety of ways or from a variety of perspectives. Earlier, I described it as a combination of comprehension and rehearsal - which is a description of learning as an external manifestation. I will return to that description shortly, but first, let us remain with an internal perspective. I want to describe learning as a recording of the regularities and of the irregularities of experience. Learning is the consignment of these to memory.

The basic stuff of learning consists of regularities in the world, while its power lies in the abstraction process provided by thought. Let's see the workings of all this. When an event recurs many times in my environment, I not only pay attention to it, it is also encoded more strongly in memory. The regularity is noticed and encoded. Our cognition is geared to searching for and dealing with such regularity. Regularity is one aspect of coherence. It is what makes it true, i.e. coherent, that the sun rises every morning.

Regularity cuts across levels of abstraction. It forms the basis of conditioning just as it underlies coherence in abstract mental structures. Thus, associating a word processing error with a particular keystroke may happen over a number of such keystrokes and lead to the avoidance of these keystrokes in a given situation. Likewise, abstractly learning that the 'del' command will delete my file assumes that it will do so every time I use the command; the regularity involved ensures coherence.

The associative processes of learning are built upon regularity. If an event recurs many times, say involving a behavior and a rewarding result, or simply involving two or more congruent elements occurring repeatedly together, then it becomes encoded. The individual does not actively encode the event; rather, it gets encoded through the regularity that happens to the individual. Here, learning is a process activated by a context involving regularity, and the process leads to an association in the memory structure of the individual.

There is also another approach to learning, one initiated this time by the learner. The same basis in regularity is involved here too. Suppose you want to learn some Italian vocabulary, say the terms for city, town, and so on. These will not occur by themselves in your environment in a regular enough manner for you to learn them. So you practice them voluntarily, i.e. you make them occur regularly through rehearsal. Rehearsal is nothing more than the self-initiated regularity-generation of events, for instance the event 'city-citta'. Thus, both natural occurrences of regular events and artificially generated ones lead to associations in memory. The detailed parameters of this process form much of what is studied by cognitive psychologists.

Now let's look at abstraction. It was said earlier that abstraction provides power to learning. Not surprising, given that it is the basis of thought. Consider how it might assist learning. Regularities in experience can simply be encoded in a brute manner in memory, as happens in conditioning. Alternatively, the regularities can be noted by the individual and labeled, i.e.. composed into a concept. A concept is but an abstract element that embodies a regularity. Its power lies in the fact that it can then be used for thinking, i.e. it is a handle on a complex of experiences that themselves are unwieldy. However, summarized in a concept, they become manipulatable. Building a concept is abstracting the essence of a regularity out of the detailed experiences and giving it a handle, i.e. a label.

Concept building is probably rare, though. More usually, we start from the label (the word dog, for instance), and associate to it an element of experience that recurs over many occasions (seeing many dogs, while having the word said). The associated processes of generalization and discrimination are but tuning of concepts, and must also involve regularity.

Once a group of concepts are developed, they can be manipulated not only for problem-solving, but also for further learning. This forms an interesting situation, since it by-passes regularity, which we considered to be the basis for learning. Concepts can eventually be verbalized in terms of their abstract elements: a dog is an animal of a certain size and appearance, that has a certain disposition and behavior, and so on. Thinking now enables one to juggle these features and derive further concepts, usually through contrast and analogy: thus, a wolf is like a dog, except that it lives in the wild and has a particular behavior. What is happening in terms of learning is the following: the regularity is still there, only it is hidden behind the meaningful features of the concepts. It is surreptitiously used during the thinking that goes on in the derivation of new concepts. Thus, instead of experiencing the regularity of wolves in the wild, one experiences the abstract features of this regularity. Abstraction enables one to play games with the concepts themselves, rather than to play directly with tangible elements of the physical world around us.

In summary, it can be said that regularity in the world, perceived either directly through experience or abstractly through concepts, forms the stuff of learning and results in associations in the mind. Some of these associations are between simple elements in the world, such as those involved in conditioned responses or in arbitrary regularities like names for things ('dog' and not 'dag' for instance). Other associations are between abstractions (democracy and government) or between abstractions and arbitrary elements (the concept of democracy and the US Declaration of Independence). Now let's look at the central role of coherence in our abstract learning.

COHERENCE

Coherence is the logical way in which things fit together. A quick example: I am told that Kiev is the capital of the Ukraine and I know that a political entity like a country only has one capital, so Kiev is it; I won't go around looking for a second one and so on. Logic and the implications derived from it constrain thought (thereby making it possible, it should be noted), which facilitates the cognitive processing required for meaningful learning. Without the constraining factor of logical thought, everything would have to be learned in a purely associative manner, which would put a terrific strain on the cognitive system and severely limit the capacity for cognition.

What learning involves beyond the associative level is a constant search for coherence in the information provided to us by our environment. Understanding is the grasping of this coherence, whereas lack of comprehension is the failure to derive a coherent representation.

Coherence applies to learning in two different ways. First, the information encountered must be internally consistent. For instance, if a textbook states that the brain is composed of three components but then goes on to describe four of them, that is inconsistent and will lead to difficulty in comprehension. The information, in that case, would lack coherence. Second, the information encountered must fit with the learner's current cognitive structure. A novice in a domain cannot grasp complex technical information simply because the basic mental models that would be needed are yet to be acquired.

Cognitive structure has been an important focus for thinking within cognitive educational psychology and it is useful to see why. On the one hand, it has practical import for instruction: the recognition that information must fit into an existing mental structure focuses attention on ensuring that there is a potential match, and if not, doing something about it, such as providing some organizing information in advance of the more specific information. And on the other hand, it embodies the structural aspects of mental models, which are the end result of learning efforts.

However, we are dealing here not so much with the conditions of learning as with the process itself. Learning, then, is a search for the coherence that lies within the information that we encounter. If the search fails, then we either do not learn the information, er else we attempt to learn it by rote, that is we fall back on an associative rehearsal-based approach to learning. But if the search for coherence succeeds, we attain understanding, which constitutes a successful learning experience.

There is an interesting relationship involved here: if understanding comes with some difficulty, the information will be well-remembered, but if understanding is too easy, the memory trace will be less robust. A practical illustration of this comes from the discovery approach to instruction (as contrasted to expository instruction): discovery learning may be long and arduous, but once understanding is achieved, the insights generated are memory-robust.

This relationship is based on the coherence principle discussed above. Difficulty in initially understanding will lead to a search for coherence that will involve the exploration of a number of avenues, and when understanding is achieved, it will be because a proper coherence has been perceived. The effort after meaning that such exploration involves is not the crucial ingredient in the process, but merely an external manifestation of it. It is the achieved coherence that is at the heart of learning.

But why could learning, in terms of eventual memory for the information, suffer from understanding being too easy? The answer lies not directly in the process of achieving coherence, but more directly in the degree of coherence involved, and, congruently, in the non-coherent information present. If understanding is easy, it is likely that the required cognitive structure for the information being considered is already well established, i.e. it does not need learning itself. What is being learned then in such a situation is mostly factual information, needing associative processing, and not structural information requiring a search for coherence. As an illustration, consider the learning of general biology. It is a field that is not generally difficult to understand because we have the general mental structures for discussing its elements. It is a field, however, in which a great deal of particular information of an associative type must be learned, especially terminology. Thus, coherence requirements are limited, and learning focuses instead on associative processes.

In summary, coherence lies at the very center of the process of learning. In as much as information coheres together to form a structure, it leads to understanding and to easy memory retrieval later on. But, in as much as information does not cohere together, i.e. in as much as it is arbitrarily defined, it must rely on the replicability of associative learning to be later well remembered.

A great deal of our symbolic world is arbitrary, especially our language. That so and so is called John has no reason to it, for he could have been called Peter, Paul, or any number of other names. Likewise for terminology: that a class of animals is called marsupials, or another dogs, cats, and so on is completely arbitrary. It is true that some terms are derived from others, and therefore not quite arbitrary (for example, microscope from micro and scope), but these terms are few within the full set of terms that we use. Many other elements of information are arbitrary, for instance that Jupiter has 16 known moons and not 12, or that the largest ocean is the Pacific and not the Atlantic. These are all elements of information for associative learning.

Now, the constraining force of logic makes certain elements of information non-arbitrary. If the telephone on my desk is gray, then it is not yellow, red, or blue; indeed, under normal circumstances, things do not change color. This is a logical constraint in the world. Notice that I am dealing here with a practical logic, one that generates practical implications, and not with formal logic that settles quibbles in rationality. It is this practical logic that will tell me that, if my phone is broken, it will not work again until I have it fixed. Logic simply constrains the world of possibilities, thereby reducing the arbitrariness of some of the world we encounter.

This reduction is arbitrariness not only enables us to creatively think by symbolically playing with the world, but also reduces the need for arbitrary associations, and hence reduces the load on learning. The reason is simple: we also use our practical logic in recalling the information that was learned, i.e. in employing the fruits of our learning for whatever purposes are required at the moment. Whereas logic can be no help at all in attempting to recall arbitrary associations such as the fact that the capital of Austria is Vienna, it can be very helpful in regenerating information from bits and pieces in memory. I can actively use my powers of logical thinking; for instance, if I am having trouble recalling a symptom-fault relationship in an electronic system, I can use my schematic tracing skill to attempt to replicate the fault by following its logical pattern in the schematic. It is in this sense that knowledge builds on knowledge.

DECLARATIVE AND PROCEDURAL KNOWLEDGE

A distinction often made in discussions of knowledge and learning is the one between declarative and procedural knowledge. Declarative knowledge is knowledge of what is, whereas procedural knowledge is knowledge of how to do something. The distinction, however, has little practical value for a theoretical analysis of learning, even if it has proven useful in the area of knowledge representation in artificial intelligence.

There is nothing special about procedural knowledge in terms of learning. It is true that it differs from declarative knowledge by being concerned with personal human actions (our doing something), a class of learning results that we have come to call skills. It thus involves us, personally, in what is learned. I am not learning about the world, but rather, I am learning about acting in the world, about having an effect on the world. Thus, the object of learning is different. But what about the process?

Other than the fact that skill learning can involve motor actions (also a question of what is learned), there is no fundamental difference in the process of learning. We are still dealing, just as in learning declarative knowledge, with associative and structural processes of learning. Condition-action relations are established, as well as sequences of actions, all of these elements eventually forming a full complex skill. Learning these elements is subject, like all learning, to repetitive encounters to establish the regularity, and it can be facilitated during the process by structural processing. Thus, for instance, learning to drive a car involves establishing the relations and sequences of action, and then practicing the skill until it is well established, i.e. repetitively engaging the actions to establish the proper associations.

Now, there are two concerns that crop up with respect to skills: that of getting it right and that of performing it quickly. Getting it right often involves a lot of first getting it wrong and then gradually shaping and fine-tuning the behavior until it is indeed right, or at least acceptably so. This, however, is no different than committing a poem to memory. In both cases, there are errors, found out to be such from the feedback provided by the world, and eventual latching onto the correct behavior as practice proceeds. Thus, learning the capitals of Europe and learning how to tie a bow-tie are not different in how learning proceeds.

The other issue is that of automatically. Skills eventually get automaticized, often to an extent that we no longer pay attention to actually doing them, and indeed, that we can hardly say anymore how we do them. Knowledge becomes so compiled as to be inscrutable, yet it is extremely effective. Practice of sequential actions leads to such robust associations that the sequence becomes unitized into a composite skill. The process is somewhat akin to that of conceptual abstraction-making, in which the individual elements of experience unitize into abstract concepts that now represent a whole class of elements.

With procedural knowledge, super-practice makes the action sequences very quick and efficient, because that is what is needed. We rarely need declarative knowledge to be so quick, so we don't practice it as much for the sake of speed. But we could, if we had to, or if we wanted to. We actually do something of the sort when we practice basic arithmetic, such as 5+5. At first, children have to think about it a lot, they come up with false starts and correct themselves, and they rehearse a lot. Eventually, the answer comes naturally, without effort, just like driving a car.

Some of the examples I used above may seem a little strange: after all, reciting a poem or providing a sum do seem rather procedural, even if they involve not so much the personal relationship of the individual with the world as they involve things about the world. The fact is that all we learn, about ourselves or the world, involves performing and thus involves doing something. Just as procedural knowledge can be seen declaratively as actions in relation to the world, so too can declarative knowledge be seen as the individual producing actions in a procedural fashion, even if the actions involve no more than reciting something or recalling something. In a sense, no knowledge is impersonal, and that is why it cannot simply be poured into the mind, despite the analogy with electronic information, which can be copied at will from one computer knowledge base to another.

HANDLING INFORMATION

Our cognitive processes are tuned to the handling of vast quantities of information of all kinds, be it enactive, iconic, or symbolic information. In our dealings with the world, we seek to establish regularities that will have predictive value and we seek to infer from the information given other consequences that logically entail from what is at hand. As in perception, we seek cognitive closure that leads to a tidy world where things fit well together. Our attention is thus drawn, once regularities are established, to irregularities in the world, and to information in the world that does not seem to fit the logical picture. That is why the bizarre always stands out, not just socially, but cognitively as well. The odd element in a series will be remembered, whereas the others blend into a hodgepodge. The incongruity in a logical system will have us trying to work out a way to understand it. This is to a large extent what forms the basis of curiosity.

Thus, the regularities of the world and our capacity to juggle them in thought form the basis for all learning, be it a learning that comes naturally from occurrences in the world, or learning that we must help along by structuring events such as rehearsal activities of various sorts.

There is another aspect of our handling of information in our exchanges with the world, and that is our activity cycle. It involves a push towards exploration into the unknown, but also a pulling back in order to take stock and consolidate the information encountered. We are driven by curiosity to continually seek out new information, but we also continually come up against the limits of understanding. So we also continually seek to incorporate what we encounter into our current cognitive structures. We are ever involved in an effort after meaning. On the one hand, we are pushed towards novelty, but on the other, we are held back by the need to sort out the novelty and make it ours.