Introduction

This is a course about learning and memory -- but mostly, frankly, about memory. The two topics are, obviously, closely related. Learning could not take place without the capacity to retain what has been learned. And nothing gets retained in memory unless it has been learned.

Learning

The traditional definition of learning is as

a relatively permanent change in behavior which results from experience.

This definition rules out innate responses to stimulation, such as:

reflexes, innate responses to an environmental stimulus that involve some discrete portion of the somatic or autonomic nervous system, such as the patellar reflex;
taxes, gross responses to an environmental stimulus (such as orienting, or moving toward or away from the stimulus) that involve the organism's entire skeletal musculature, such as the positive phototaxis of a moth to light;
or instincts (otherwise known as fixed action patterns), species-specific responses to particular patterns of stimulation that are universal within the species, such as the "zig-zag dance" of the stickleback fish.

It also rules out more permanent behavioral changes caused by other processes:

maturation, or the unfolding over time of a genetic program, as when an organism "learns" to walk or fly;
direct alterations of anatomy or physiology, produced by lesions, surgery, or drugs.

From a cognitive point of view, however, learning is better defined as

a relatively permanent change in knowledge that results from experience.

The canonical cases of learning are the classical conditioning discovered by the Russian physiologist I.P. Pavlov (who got the Nobel Prize for his trouble), and the instrumental conditioning discovered by the American psychologist E.L. Thorndike -- both in 1898. (The terms themselves were popularized, if not coined, by Hilgard and Marquis (1940).

In classical conditioning, a conditioned stimulus (CS) is paired with an unconditioned stimulus (US), and the CS comes to elicit a conditioned response (CR) usually resembling the unconditioned response UR. The knowledge concerns the organism's ability to predict events in the environment -- to know, or at least to expect, that the conditioned stimulus will be followed by the unconditioned stimulus.
In instrumental (operant) conditioning, a CR is reinforced in the presence of a CS, and the CR becomes more likely to occur in the presence of the CS. The knowledge concerns the organism's ability to control events in the environment -- to know, or at least to expect, that some behavior will be followed by some consequence.

But there are also special cases of learning.

In language acquisition there is a change in behavior resulting from experience, in that the child learns to speak and understand his or her native language via exposure to it. This fits the definition of learning, although the learning is not mediated by the contingencies of reinforcement (though this point, gospel to Chomskian linguists, has been disputed by later connectionist theorists).
Similarly, in perceptual learning, as in visual adaptation to distorting prisms. Again, it is not clear that conditioning processes are involved.
Finally, there is observational learning. Arguably, most human learning is not mediated by direct experience of stimulus, response, and reinforcement. Rather, most human learning occurs through vicarious experience:

by example, through observing the behavior and experiences of others; and
by precept, through deliberate, sponsored teaching by others.

There are examples of observational learning in nonhuman species -- most prominently, in Mineka's classic experiments on the observational learning of fear in rhesus monkeys. According to the cognitive social learning theory promoted by Bandura and others, observational learning may have special rules.

No matter what kind of learning is involved, whatever has been learned obviously has to be stored somewhere, so that it can be used later as appropriate. Without a capacity for memory, organisms would have no ability to permanently change behavior as a result of experience.

Memory

Given the cognitive definition of learning as a change in knowledge resulting from experience, as a first approximation we may define memory as the mental repository of stored knowledge, acquired through experience -- including knowledge of the learning experience itself.

The concept of memory could be expanded to include innate knowledge, which is a product of the organism's evolutionary and genetic heritage. But we have to draw a line somewhere -- and we want to draw it before we get to reflexes, taxes, and instincts. Thus, in this course, we will limit memory to mental representations of knowledge acquired through experience.

Even so, it's clear that the contents of memory can range widely. Memory can take many different forms, corresponding to the many different forms that knowledge can take.

One hierarchical classification of memory begins with the distinction between two forms of knowledge -- a distinction with origins in computer science (Winograd, 1975) that was imported into psychology (Anderson, 1976):

Declarative knowledge consists of factual knowledge, or knowledge with truth-value, which can be represented by sentence-like propositions consisting of a noun phrase, a verb phrase, and another noun phrase. Thus:

Automobiles have wheels.
Columbus discovered America.
The hippie touched the debutante.
Last night I had chicken for dinner.

Procedural knowledge consists of directions for action, either overt (behavioral) or covert (mental), which can be represented as "if-then" productions consisting of a goal, a condition, and an action. Thus:

If you want to multiply 3 x 2, then add 3 to itself one time.
If you want to drive a standard-shift car, then press down on the clutch.

The knowledge acquired through classical and instrumental conditioning can be thought of as procedural in nature:

For the Pavlovian case, If bell then meat powder.
For the Thorndikian case, If you're in a puzzle box and you want to escape then press the lever.

Of course, no individual piece of knowledge stands in isolation. Each item in memory is linked to many others.

In declarative memory, there is an associative network where nodes represent concepts, and associative links represent the relations between them.

In procedural memory, there is a production system in which the output of each individual production serves as input to some other production.

And, for that matter, declarative and procedural memory are also related:

The goals and conditions of a production are represented as nodes in declarative memory.
A common action in procedural memory is to create a new node in declarative memory.

In the hierarchical classification of memory, procedural knowledge can be subdivided into two categories:

behavioral actions, in which the output of the production is some form overt behavior;
mental actions, in which the output of the production is some covert mental state.

Following Tulving (1983), we can subdivide declarative memory into two categories:

Episodic memory concerns factual knowledge about discrete events -- episodes in an organism's experience -- with each event tied to a unique spatiotemporal context (two events can't occur at the same time in the same place). Episodic memory is, in a sense, autobiographical memory, because it concerns the behaviors and experiences of an individual organism (though, as we will see, autobiographical memory is more than a record of discrete episodes).

Semantic memory concerns more abstract, generic, context-free knowledge. Semantic memory constitutes a sort of "mental dictionary" or encyclopedia of world-knowledge.

Like learning and memory, episodic and semantic memory are related.

Semantic memory develops as similar episodes are repeated.
Semantic memory forms the cognitive background against which individual episodes are encoded.

In this course, we are primarily concerned with episodic memory, but we will also have things to say about semantic and procedural memory.

Mind and Brain

Memory is a mental capacity, and episodic memories are mental representations of past events, and this course will be primarily concerned with the psychological analysis of memory as an aspect of mental life. But the brain is the physical basis of mind: whatever is represented in memory also has to be represented in the brain. Accordingly, this course will also discuss the anatomy and physiology of learning and memory -- what physiological changes underlie the changes in behavior that take place as a result of experience; and the neural representation of knowledge and experience.

Some comments are in order concerning the relationship between cognitive psychology and cognitive neuroscience. Neuroscience is a very useful tool for psychology, because it connects psychology to the other biological sciences. Some neuroscientists make the strong claim that neuroscientific knowledge will constrain psychological theory. While I would quarrel with that position, because psychology is a biological science (as well as a behavioral science, a cognitive science, and a social science!), the relations between the mental and the physical is an enduring question for psychology.

But, as someone (not me) once said, physiology is a tool for psychology, not an obligation. A scientific analysis of memory can proceed exclusively at the psychological level -- deriving the principles of how knowledge is represented, and how memories are processes. From this point of view, the biological basis of learning and memory is an interesting question, but it is not the focus of psychology.

In any event, advances in knowledge at the biological level of analysis depend on knowledge at the psychological level. Psychological theory tells neuroscientists what to look for in the brain, and shows them how to interpret brain activity. Put crudely: you can put subjects in a brain-imaging machine and give them something to learn, and some part of the brain will surely light up. But without a clear understanding of the task at the psychological level of analysis, we would have no idea what that part of the brain is doing. We would be left only with a kind of sophisticated phrenology.

Stage Analysis of Memory

Memory can be thought of as the mental storehouse of knowledge, but it is also an activity. when knowledge is acquired through learning, retained over time, and used in performance, memories are being processed.

In cognitive psychology, we analyze the processing of episodic memories in terms of three stages:

Encoding: the laying down of a trace of perceptual activity, which leaves a representation of the event in memory.
Storage: retaining a newly encoded trace in memory over time, in a latent state, able to be activated when needed.
Retrieval: recovering the encoded trace from storage, and using it in some mental or behavioral activity.

Note that, in principle, successful remembering requires success at all three stages:

The new event must be encoded.
Encoded knowledge must remain in storage.
Stored knowledge must be retrieved.

Therefore, any particular instance of forgetting can involve a failure at any of these stages, alone or in combination.

Stage analysis is reflected in the library metaphor of memory, in which a memory trace is analogous to a book:

Encoding: the book must be purchased, cataloged, and shelved.
Storage: the book must not be destroyed by fire or insects.
Retrieval: the book must be looked up in the catalog, found on the shelf, opened, and read.

The encoding stage is where learning takes place. Everything depends on this acquisition stage: no learning, no memory.

This emphasis on encoding illustrates the intimate relationship between memory and perception: No perception, no memory. Every experience changes the contents of memory.

Following Richard Semon and Karl Lashley, we can define the engram as the lasting trace of experienced objects and events, which persists in memory after the experience is over. According to the encoding specificity principle, to be discussed later, what can be retrieved depends on what has been encoded -- how an event has been processed at the time of perception.

Of course, perception also depends on memory. Jerome Bruner famously observed that every act of perception is an act of categorization. Perception is not complete until the stimulus event has been identified and categorized. In this process, the perceptual representation makes contact with pre-existing knowledge stored in memory.

In the final analysis, perception combines information from two sources:

Information extracted from the stimulus environment: the features and patterns of features in the object or event itself, which provide the stimulus basis for perception.
Information retrieved from memory: knowledge and beliefs about the world, which provides the cognitive basis for perception.

Beyond the Library Metaphor

But the library metaphor makes clear that memory is not just a matter of encoding; it's also a matter of retrieval. Stored memories are of no use unless one can gain access to them. Tulving and Pearlstone (1966) made an important distinction between the availability of a memory in storage and the accessibility of that memory at the time retrieval is attempted. Knowledge is made available by encoding and storage processes, but we gain access to available knowledge through retrieval processes. And retrieval is not simply automatic.

The library is a powerful metaphor for memory, but it is also misleading. F.C. Bartlett (1932) had the insight that both perception and memory entailed effort after meaning: in perceiving, and remembering, we try to make sense of what we perceive and remember.

Put another way, both perceiving and remembering represent problem-solving activity.

Thus, perception is not just a mater of extracting features from a stimulus, and recognizing patterns of features as familiar. In Bartlett's view, we must figure out what is happening now, making sense of the percept in terms of expectations and past experience. This "figuring out" involves "higher" mental processes of reasoning, inference, judgment, and decision-making.
Similarly, memory is not just a matter of retrieving memory from storage. The rememberer (by analogy to the perceiver) must figure out what happened in the past, making sense of the memory trace in terms of current context.

So, just as perception is constructive in nature, so memory is reconstructive in nature. Both cognitive activities rely heavily on judgment and inference.

From Bartlett's point of view, then, we recreate events and experiences each time we remember them. Beginning with fragmentary details stored in the memory trace, and suggested by other information available at the time of retrieval (including information contained in the query itself), we then engage in reconstructive activity, filling in gaps and fleshing out details, making inferences from our world-knowledge about what might have happened.

The point of Bartlett's analysis is that remembering is not so much like reading a book -- it's more like writing a book anew from fragmentary notes.

Thus, memory is linked not just to perception, but also to other cognitive processes of reasoning, judgment, inference, problem-solving, and decision-making. Because memory depends on both perception and reasoning, it stands at the very center of cognitive psychology.

This page last modified 05/27/2014.