Introduction: What Is Learning and How Can We Study It?

After reading this chapter you should be able to:

• State the definitions of the standard three forms of learning, as well as complex learning.
• Identify the differences between learning as a procedure, effect, and mental process.
• Describe the functional and cognitive approaches in learning psychology and the relation between the two.

Introductory Task

• Describe what conditioning means to you.
• Find examples of behavior that you believe can or cannot be changed via conditioning.
• Explain how those changes in behavior come about. Which processes are the basis for those changes in behavior?
• Consider whether conditioning is relevant only for the study of nonhuman organisms. To what extent do you think conditioning also occurs in humans?

At the start of a book on the psychology of learning, it is important to reflect on what exactly the psychology of learning is all about. Like any other scientific discipline, the psychology of learning is defined by its goals. Scientists can have a multitude of goals that differ with regard to not only the object that is studied but also the questions that are asked about that object. Hence, to define an area of research, we need to specify the object of study as well as the questions that are asked about that object.

When applied to the psychology of learning, it seems obvious that the object of study is “learning.” But what exactly is “learning”? Although research on learning has a long history in psychology and is still important in the field today, there is surprisingly little consensus about the definition of learning. In reviews of the literature one can find various definitions that differ in important ways (see, e.g., Barron et al., 2015; Burgos, 2018; Lachman, 1997). On second thought, this lack of agreement is to some extent unsurprising, because it is often difficult to reach consensus about definitions. For instance, could you agree with others on the definition of a chair? Must all chairs have four legs? Is anything you can sit on a chair? Nevertheless, in science, it is important to be as explicit as possible about the definitions of core concepts, because those definitions have a huge impact on what researchers do and why they do what they do. In fact, many discussions in science can be traced back to hidden differences between researchers in the way that they define core concepts in their science (Wittgenstein, 1958). In this introductory chapter, we discuss the way we define learning. Our definition deviates somewhat from those used in everyday life and by other scientists. We do not claim that our definition is the best possible definition. Like most other definitions we present in this book, our definition of learning is best thought of as a working definition. However, we do believe that our definition is a very useful one. Compared to other definitions, it has three important qualities: precision (it delineates in a clear manner what does and does not count as an instance of learning), scope (it is broad enough to encompass a wide range of learning phenomena), and depth (it is compatible with research in other areas of science such as biology, genetics, and neuroscience). Rather than enter into unproductive debates about whether our definition captures the “true” essence of learning, we hope that this book will help readers to appreciate how useful our definition can be.

Figure 0.1

Let’s start!

Another advantage of our definition of learning is that it is compatible with two fundamentally different approaches in learning research (and psychology in general). First, the functional approach is concerned primarily with describing the factors that moderate learning (when does learning occur), and more specifically, with distilling abstract principles of learning from concrete findings. Intellectual traditions such as radical behaviorism, behavior analysis, and more recently, contextual behavioral science, draw heavily on the functional approach; we will return to this later in the book. Second, the cognitive approach aims to describe the mental mechanisms that mediate learning (i.e., how does learning occur). Although both approaches have been well established for many years, there is little interaction between them. In fact, most textbooks on the psychology of learning focus on just one of these approaches (either a functional or a cognitive approach to learning). In this introductory chapter, we not only try to capture the core of each approach but also—aided by our definition of learning—clarify that both approaches are compatible and mutually supportive. Hence, we provide a unique functional-cognitive framework for learning research that reconciles these two main approaches in the psychology of learning.

As the result of addressing these complex issues, this introductory chapter has become quite long, much longer than what is typical in textbooks about learning. Although we illustrate our ideas with concrete examples, some of the points we make remain quite abstract, and some of you may struggle with this introductory chapter. It is, however, our firm belief that a full appreciation of the studies and theories covered in the subsequent chapters requires a coherent perspective on the nature of learning as well as the different questions about learning that psychologists pose and try to answer. We therefore hope you will follow us in this important first step of our journey.

0.1   What Is Learning?

0.1.1   Learning as Ontogenetic Adaptation

Darwin’s theory of natural selection is without doubt one of the most important scientific insights that our species has ever achieved. Living creatures are not static entities; they are constantly evolving in response to the world around them. The driver of evolution is adaptation to the environment: the more an organism is able to adapt to the environment in which it lives, the greater its chances of reproducing. This means that those characteristics of the organism that improve its adaptation to the environment are more likely to be passed on to the next generation, compared to characteristics that do not improve adaptation, or even hamper it.

Darwin’s theory of evolution via natural selection focuses mainly on phylogenetic adaptation—that is, the adaptation of animal species to their environment across generations. In contrast, learning psychology can be seen as the study of ontogenetic adaptation, which is the adaptation of individual organisms to the environment during the lifetime of those organisms (De Houwer, Barnes-Holmes, & Moors, 2013; Skinner, 1938, 1984). Learning can thus be defined as an observable change in the behavior of a specific organism as a consequence of regularities in the environment of that organism. In order to say that learning has occurred, two conditions must be met:

• An observable change in behavior must occur during the lifetime of the organism.
• The change in behavior must be due to regularities in the environment. Learning is thus seen as an effect—that is, as an observable change in behavior that is attributed to an element in the environment (a regularity in the environment).

The above definition clarifies that learning psychology has an essential role to play in understanding the behavior of all organisms—human and nonhuman alike. In much the same way that a species will adapt to the environment across different generations, so too will individual organisms adapt to the environment during the course of their own lives. The goal for learning psychologists is to arrive at the best possible understanding of ontogenetic adaptation. Before we consider how this might be achieved, let us first consider in more detail the implications of our definition of learning.

0.1.2   Difficulties in Applying the Definition of Learning

Although the above definition certainly helps clarify what we mean by learning, there are still several issues to consider when determining whether learning has taken place. First, recall that the observed change in behavior (due to regularities in the environment) can occur at any point during the lifetime of an organism. For instance, the impact of a regularity on behavior might be evident immediately, or only after a short delay (e.g., one hour), or even after a long delay (e.g., one year). It is therefore difficult to conclude with certainty that learning has not taken place when a regularity is present but there is no change in behavior, because it is possible that a change in behavior will only occur at a future point in time.

Second, there may be disagreement about what is meant by behavior and thus what constitutes a change in behavior. We adopt a broad definition that includes any observable response, regardless of whether that response is produced by the somatic nervous system (e.g., pressing a button), the autonomic nervous system (e.g., saliva secretion), or neural processes (e.g., electrical activity in the brain). The concept of behavior also refers to responses that in principle are observable only by the organism itself (e.g., a conscious mental image or thought).¹

The third and perhaps most important issue is that applying the definition of learning requires that we make a causal attribution. Think back to our definition of learning as an effect (i.e., a change in behavior due to regularities in the environment). It is not enough that we merely observe a change in behavior. Rather, a change in behavior qualifies as an instance of learning only when it is caused by a regularity in the environment. There is therefore an assumed causal relation between environment and behavior built into the definition itself. We say “assumed” because causal relations cannot be observed directly. They can only be derived from either logical arguments or empirical evidence.

To illustrate, consider newborn children who initially show a grip reflex during their first six months of life (i.e., they quickly close the palm of their hand whenever it is stimulated) but then stop doing so by their first birthday. This reflex is clearly a behavior (i.e., it is a response to the stimulation of the hand). Moreover, the reflex changes: stimulating the infant’s hand (the stimulus) initially leads to a grip reflex (behavior) and this reflex gradually disappears across time. But does the change in behavior qualify as an instance of learning? It would if the change in behavior is due to a regularity in the environment. For instance, the reflex might weaken as the result of repeatedly experiencing stimulation of the palm, like when adults become accustomed to a noise that they frequently hear (e.g., traffic outside your window). Saying that the change in the grip reflex is an instance of learning thus boils down to a hypothesis about the causes of the change in the grip reflex—that it is due to a regularity in the environment (i.e., the repeated presentation of a single stimulus).

An alternative hypothesis is that the grip reflex decreases because of the spontaneous maturation of the infant’s brain. This “maturation” hypothesis maintains that during the first months of life, new neural connections are formed as a result of genetic factors that have little to do with the infant’s constant interactions with the environment. These spontaneously formed neural connections grow in strength so that, by a certain age, they inhibit the grip reflex entirely. In this case, the change in behavior is due to genetically determined maturation instead of regularities in the environment and would therefore not qualify as an instance of learning. But which hypothesis, learning or maturation, is correct? The answer ultimately depends on empirical evidence. The “learning” hypothesis would predict that a decrease in palm gripping will depend on the environmental regularity (i.e., the frequency with which the infant’s palm is stimulated). If this prediction is confirmed, then the learning hypothesis would gain support. The “maturation” hypothesis, on the other hand, would predict that the reflex will disappear as a function of time rather than due to the frequency of palm stimulation.

The key point here is that there are causal assumptions at the core of the definition of learning. This is true for other concepts as well. Take the idea of a traffic-related death. Suppose that a driver has a fatal heart attack while driving his car on the motorway. After he dies, the car suddenly comes to a stop and a second car collides with the first, and the second driver also dies. In this situation it is likely that only the second driver will be considered a traffic-related death because the first did not die due to a traffic-related factor. However, if evidence emerges showing that the second driver also suffered a fatal heart attack before he collided with the first car, then a new debate can take place about whether the death of the second driver also qualifies as a traffic-related death. Yet, even then, the classification of what caused the death ultimately depends on what we would consider a direct cause of death (e.g., was the second driver’s heart attack due to the sight of the first car)? What this example illustrates is that the ultimate criterion for determining a traffic-related death is not an objective characteristic of the situation (e.g., did the driver have a heart attack; did the death happen in a traffic-related situation). Rather, it is based on how we determine what is a direct cause of death (e.g., did the driver die due to a medical condition that was present prior to the traffic situation, or was the heart attack a consequence of the traffic situation). Sometimes the cause of something is easy to determine (e.g., the fact that flicking a switch turns on a light). But at other times there is a reason for doubt, and more research is needed. The same is true when we want to determine whether a change in behavior is an instance of learning.

0.2   What Are the Different Types of Learning?

0.2.1   Types of Regularity in the Environment

If learning refers to ontogenetic adaptation, then the goal of learning research is to study the impact of environmental regularities on behavior. But what exactly is an environmental regularity? In line with De Houwer et al. (2013, p. 634), we define a regularity as “all states in the environment of the organism that entail more than the presence of a single stimulus or behavior” (also see De Houwer & Hughes, 2017). Critically, different kinds of regularities can be distinguished. For instance, we can say that a regularity is present if (a) one particular event is repeated across time or space, (b) different events occur together at one place and moment in time, and (c) different events occur together in multiple places or moments in time (De Houwer et al., 2013).² Regularities can also differ from each other with respect to the nature of the events involved. For example, some events involve only stimuli whereas others involve stimuli and responses. We can sometimes even detect regularities in the occurrence of regularities (so-called metaregularities). The key point here is that there are many different types of regularities. Yet, strangely, the vast majority of research in learning psychology has tended to focus on one of the following three types:

1. Regularities in the presence of a single stimulus: Is a certain stimulus (e.g., food) presented, and if so, in what way (e.g., how much food, how often does it occur, etc.)? This regularity refers to the repeated presentation of a single stimulus in ways that are independent of other events such as the presence of other stimuli or responses.
2. Regularities in the presence of multiple stimuli: Is there a regularity in the presence of one stimulus and the presence of another stimulus (e.g., do you find certain foods in certain places)? This type of regularity is most often instantiated by repeatedly presenting two stimuli together in space and time. Note, however, that even the one-time occurrence of two stimuli together qualifies as more than one stimulus at a single moment and can thus be regarded as a regularity. We therefore use the term regularity in the broadest sense of the word (for more on this, see De Houwer et al., 2013; De Houwer & Hughes, 2017).³
3. Regularities in the presence of a stimulus and response: Is there a regularity in the presence of a stimulus and a response (e.g., do you get a specific beverage whenever you press the button of a vending machine)? This regularity is concerned with contingencies between stimuli and behavior.

Think It Through 0.1: Are Other Definitions of Learning Also Possible?

Given that definitions containing causal attributions are sometimes difficult to apply, the question arises: Are there other definitions of learning that are less problematic? What do you think? Try to come up with your own definition of learning and compare its usefulness with that of the definition we offer in this chapter. (For this and all subsequent “Think It Through” questions, some reflections on possible answers can be found at the end of the book.)

Think It Through 0.2: The Interaction between Learning and Genetics

Often there is a close interaction between genetic factors and regularities in the environment with regard to their impact on behavior. A good example of this is the phenomenon of imprinting. A young bird will develop a singing pattern only if it hears the vocal pattern of a conspecific during a particular period of its life. If the bird does not hear a vocal pattern, or if it hears the song pattern of a different kind of bird, or the song pattern of a conspecific outside the crucial period of its life, then it will never show the pattern later on. It seems clear that both genetic factors (e.g., for determining the crucial period) and regularities in the environment (e.g., repeatedly experiencing a certain song pattern) play a role in this particular behavior. It is as if the animal is genetically predestined to receive a print (imprint) of the song pattern of its kind during a certain period of its life.

What do you think? Is imprinting a form of learning?

0.2.2   Types of Learning

If we define learning as a change in behavior due to regularities in the environment, then we can distinguish between different types of learning based on the regularity that is responsible for that change in behavior.

1. Changes in behavior that are due to a regularity in the presence of a single stimulus (regardless of the ways in which that stimulus is potentially related to the presence of other events). To illustrate, imagine that you hear a loud bang. The first time you hear the bang, you may show a strong startle reaction. The second time you hear the bang, your startle reaction will probably be weaker. This response diminishes in magnitude even further with the third bang, and so on. So in this example, there is a change in behavior (the intensity of the startle reaction elicited by the bang declines) that is due to the repeated presentation of a single stimulus (the loud bang).
        In this book we will refer to this type of learning as a change in behavior that is due to noncontingent stimulus presentations. The term noncontingent refers to the fact that the repeated presentation of a stimulus has an impact on behavior regardless of its links with other events. The example of the reduced startle reaction to the loud bang caused by the repeated presentation of the bang is just one subtype of effects of noncontingent stimulus presentations—specifically, a subtype known as habituation (for more on this, see chapter 1)
2. Changes in behavior due to regularities in the presence of multiple stimuli (at one moment in time or across several moments in time). Suppose that your computer screen flickers (one stimulus) and a moment later you hear a loud bang coming from the computer’s speakers (second stimulus). Afterward, the screen flickers for a second time. The first time your computer screen flickers, you may do no more than look on strangely at what is happening. But the second time the screen flickers, you will probably react very differently (e.g., you may be somewhat irritated or anxious). It is possible that your reaction to the flickering of the screen changes because the flickering was previously accompanied by a loud bang. This is one instance of a type of learning known as classical conditioning. This type of learning is sometimes referred to as “Pavlovian conditioning” after Ivan Pavlov, the researcher who first studied it (Pavlov, 1927). Pavlov showed that repeatedly pairing the ringing of a bell and food leads to an increase in salivation when the bell is heard.⁴
3. Changes in behavior that are due to regularities in the presence of a stimulus and response (at one moment in time or across several moments in time). Think back to our previous example of the loud bang that comes after your computer screen flickers. Imagine that after the screen flickers (stimulus) for a second time, you (with some trepidation) press the escape key (response) and the loud bang does not follow (consequence). Chances are that when the screen flickers a third time, you will immediately press the escape key. The probability and speed with which you press the escape key increases because acting in this way reduces the chance of the loud bang occurring. In this case, the change in behavior (increase in button pressing) is due to the relation between that behavior and a certain stimulus (the presence or absence of the loud bang). This is one example of a type of learning known as operant conditioning.
        The term operant, derived from operation, refers to the fact that this type of learning is concerned not so much with the relation between stimuli and a specific behavior (e.g., pressing the escape button with your left index finger) as with the relation between stimuli and a class of behaviors that all involve the same operation (e.g., pressing a button, regardless of whether this happens with the left index finger, nose, or with a stick). Another frequently used term is instrumental conditioning. The term instrumental refers to the fact that the operation is an instrument for obtaining a certain result. This type of learning is often related to the work of B. F. Skinner, who showed, for instance, that a rat will press a lever more often if doing so is followed by the delivery of food (see Skinner, 1938).

As we mentioned previously, most work in learning research has focused on changes in behavior that are due to one of these three regularities. These three types of learning can be considered “simple,” given that they are due to a single regularity. Critically, however, other more complex forms of learning can also arise. We use the term complex learning to refer to changes in behavior that are due to the joint impact of multiple regularities. We distinguish between two types of complex learning: moderated learning and effects of metaregularities. Moderated learning refers to situations in which behavior changes as the result of multiple standard regularities—that is, regularities that have individual stimuli and/or responses as elements. In metaregularities, on the other hand, multiple regularities are embedded—that is, one regularity is an element within another regularity. Hence, standard regularities differ from metaregularities in that only the latter have regularities as elements. We will discuss both types of complex learning in detail in chapter 4. However, to give some idea about what complex forms of learning might entail, we briefly discuss a few examples in the following paragraphs.

First, sensory preconditioning (e.g., Seidel, 1959) can be seen as an instance of moderated learning (see figure 0.2). Think back to our previous example of the computer screen flickering and this being followed by a loud bang from the loudspeakers (Time 2). Now imagine that at an earlier point in time (Time 1), the same screen had also flickered and that this initial flickering was not followed by a loud bang (e.g., because the speakers were not switched on). But the flickering of the screen at Time 1 did occur right after a USB stick was plugged into the computer and its light turned on. Now imagine that later that same day (Time 3), the USB stick is reconnected to the computer and suddenly lights up. It may be that the light on the USB stick elicits fear. In the language of learning psychology, this change in behavior (increase in fear) is due to the combined effect of two regularities: (a) the co-occurrence of the illuminated USB stick and the flickering screen at Time 1 and (b) the co-occurrence of the loud noise and flickering screen at Time 2. Both regularities are standard regularities because they involve only individual stimuli as elements. Because the change in behavior is produced by the jointed impact of these standard regularities (i.e., neither regularity alone would produce the change in behavior), it qualifies as an instance of moderated learning.

Figure 0.2

Sensory preconditioning: Behavior changes at Time 3 as a result of the joint impact of the standard regularities at Time 1 and Time 2.

Second, one way of studying effects of metaregularities is by means of a procedure called the relational matching-to-sample task (e.g., Ming & Stewart, 2017; see figure 0.3). On each trial of this task, participants encounter three pairs of stimuli on a computer screen. Each pair consists of stimuli that are either identical (e.g., 3–3, 1–1) or different to each other (e.g., 6–7, 5–8). The stimulus pair at the top center in figure 0.3 is called the sample pair, whereas the two pairs of stimuli at the bottom are known as the comparison pairs. Participants must choose one of the two comparison pairs at the bottom based on whether it matches the sample pair at the top. In the example depicted in figure 0.3, there is a “match” when the relation between the sample pair stimuli (i.e., the two digits are identical or different) corresponds to the relation between the comparison pair stimuli. For instance, in the situations depicted in figure 0.3, one must choose the pair at the bottom left of the left-hand trial (because both 3–3 and 1–1 consist of two identical digits) and the pair on the bottom right of the right-hand trial (because both 5–8 and 6–7 consist of two nonidentical digits).

Figure 0.3

Example of two trials during a relational-matching-to-sample task.

If you think about it, each of the stimulus pairs in the relational matching-to-sample task can be regarded as a single regularity (i.e., they involve two stimuli being paired with each other). There is also a regularity in the occurrence of these regularities because different pairs co-occur. Moreover, this pairing of pairings is, in its turn, part of another regularity in the way that participants are rewarded for selecting comparison pairs (i.e., the match between the sample and comparison pairs indicates which comparison pair should be selected). The fact that people are able to systematically select the correct response in a relational matching-to-sample task can thus be seen as one example of an effect of metaregularities. We will return to these and other examples of complex learning chapter 4. For now, it is important only to be aware of the fact that there are complex forms of learning that involve the joint effect of multiple regularities.

0.2.3   Difficulties in Determining Different Types of Learning

Just as it is not always easy to determine whether a change in behavior is caused by a regularity in the environment (and is thus an example of learning), so too, is it not always easy to determine which type of regularity was responsible for a given change in behavior (and thus determine what type of learning has occurred). In daily life, different regularities are often present simultaneously. In such situations, it can be unclear which regularity is responsible for which change in behavior.

To illustrate this, let’s return once more to our example of the flickering computer screen. Imagine that we observe a change in a person’s reaction to the flickering screen: the first time the screen flickers, their reaction is neutral; the second time, their reaction is more negative than before; the third time, it’s even more negative, and so on. Earlier, we assumed that this change in behavior (increased negativity of the reaction to the flickering screen) was due to the regularity involving the flickering of the screen and the loud bang (i.e., stimulus pairings). However, there is another possible explanation. The increase in negativity toward the screen occurs in parallel with the repeated experience of the flickering screen. Thus in principle it is possible that the mere repeated experience of a slightly irritating stimulus such as a flickering screen can result in an increase in the irritation that the stimulus elicits (regardless of that stimulus’s relation to other events in the environment, such as the loud bang). In this way, the change in behavior (increased negativity) could be an example of the first type of learning we discussed earlier (i.e., the effect of the repeated presentation of a stimulus) or the second type of learning (i.e., the effect of stimulus pairings). It is also possible that repeatedly experiencing a loud bang causes more negative reactions to the flickering screen, independent of the relation between the loud bang and the flickering screen. So how do we know which type of learning it is? As is often the case, additional research would need to be carried out. For instance, such work could check to see if the negative reaction to the flickering screen still changes if it is experienced repeatedly without the loud bang. If so, then there is support for the idea that it is an instance of the first type of learning (i.e., effects of noncontingent stimulus presentations). If there is an increase in negativity only when the flickering screen is followed by the loud bang, then support grows for the idea that it is an instance of the second form of learning (i.e., classical conditioning).

Although there are ways to identify the causes of behavior in daily life (e.g., by using so-called single case designs), it is often practically difficult to conduct the research that is necessary to discover those causes. Learning researchers therefore often turn to experimental research in the laboratory. After all, in the laboratory they have far more control than in the external world over the regularities that a person experiences. One of the main reasons learning research is an experimental science (i.e., a science that uses the experimental method) is because experimental control is necessary in order to determine whether and what type of learning has occurred. However, it is very important to distinguish between learning effects and learning procedures. Procedures are what the researcher does in creating an experimental context in which a certain regularity is present and a certain behavior is observed. Depending on the type of regularity that is being studied, one can therefore refer to different types of procedures, such as procedures with noncontingent stimulus presentations, classical conditioning procedures, and operant conditioning procedures. In contrast to an effect, a procedure is something objective; it is no more than a list of objectively observable actions that the researcher carries out when conducting an experiment. An effect, on the other hand, implies a nonobservable causal relation (e.g., the assumption that a certain regularity was responsible for a change in behavior). What is important to appreciate here is that one cannot determine which form of learning has occurred simply on the basis of the procedure. It is not sufficient to establish that a certain regularity is present in the procedure. To be able to speak of learning as an effect, one must also be able to argue that a regularity is the cause of the observed change in behavior. We will return to this vital distinction between procedures and effects repeatedly throughout the book.

0.3   A Functional-Cognitive Framework for the Psychology of Learning

Now that we have described the topic of learning research, we can discuss its goals. A science is defined not only on the basis of what is studied (e.g., different scientific disciplines may have the same subject) but also on the basis of what one wants to know about that subject. If we look at the history and current state of learning research, we can distinguish between two sets of researchers who have two different but related goals. The first goal is to describe the environmental factors that moderate learning. We call this the functional approach within learning psychology: learning depends on (and is therefore a function of) elements in the environment. Note that the word functional is thus being used in the mathematical sense of function (“X is dependent on Y”) and not in the sense of functionality (“X is at the service of Y”).⁵ The second goal is to describe the mental processes that mediate between environmental regularities and behavior. This is the goal of the cognitive approach within learning psychology. We first provide more details about each of these approaches and then discuss their relation to one another.

0.3.1   The Functional Approach within Learning Psychology

0.3.1.1   The environment as a moderator of learning The main goal of research for those who adopt the functional approach to studying learning is to understand the environmental conditions under which regularities in the environment influence behavior (i.e., the environmental conditions under which learning occurs). Functional learning psychologists achieve this by carrying out experiments in which they intervene directly in the environment of organisms in a controlled manner in order to determine whether and when regularities result in a change in behavior. In other words, they construct and manipulate different types of learning procedures. Every learning procedure consists of several factors: (a) an environmental regularity brought about by means of stimuli and/or behaviors; (b) a behavior assumed to be causally related to that regularity; (c) an organism that exhibits that behavior; and (d) a broader context in which the regularity occurs and the organism is located. Each of these factors can moderate learning; that is, they can determine whether and to what extent a regularity in the environment has an impact on behavior. We refer to these factors as potential moderators of learning⁶ and organize them into five groups:

1. The nature of the stimuli and behaviors that constitute the regularity. If you want to study the impact of regularities on behavior, then you have to choose certain stimuli or behaviors to create that regularity. In the case of noncontingent stimulus presentations, you must choose a stimulus that you are going to present. In principle, this can be anything from a very simple stimulus (e.g., a noise with a certain duration, frequency, and intensity) to a very complex one (e.g., a complete musical piece). The same is true if you want to study the impact of stimulus pairings on behavior (i.e., one will have to choose certain stimuli that will be paired). These stimuli can also range from the simple (e.g., the ringing of a bell, which is always followed by food) to the more complex (e.g., information about the food that certain people have eaten and information about the presence of allergic reactions that these individuals exhibit). Likewise, a regularity involving behavior and stimuli can be achieved only if you choose a certain behavior and a certain stimulus. And once again, the possible choices are unlimited, ranging from pressing a lever followed by the delivery of food to obtaining a diploma followed by appreciation from one’s parents.
2. The nature of the behavior used to test the possible effect of a regularity. Regularities in the environment can, in principle, be used to change any behavior that one can imagine. One can therefore observe any type of behavior to ascertain if a regularity in the environment has had an effect. The only condition on the selection of a behavior is that it is observable. And as we mentioned earlier, behavior comes in many different forms, including controlled behavior (e.g., behavior controlled by the central nervous system, such as speaking or pressing a lever), involuntary behavior (e.g., behavior controlled by the peripheral nervous system, such as salivation), neural behavior (e.g., a certain pattern of electrical activity in the brain), and even reactions that are observable only to the organism that exhibits these reactions (e.g., covert behavior such as thoughts and feelings; see note 1).
3. The characteristics of the organism whose behavior is observed. Learning can, in principle, occur in every living organism. The only precondition is that the organism is capable of demonstrating a change in behavior. For instance, an inanimate object such as a stone cannot learn because it is does not behave (i.e., it does not respond to stimuli in the environment) and therefore cannot change its behavior (as the result of either regularities in the environment or any other potential cause of changes in behavior). On the other hand, most living organisms, from the smallest single-celled organism to humans, do behave in ways that change. Hence, living organisms can function as objects in the scientific study of learning (see Roche & Barnes, 1997, for an insightful discussion of the concept of “organism” and its role in the study of learning and behavior). Procedures can differ with respect to the characteristics of the organism being studied. Note that moderation of learning by the nature of the organism is concerned not just with differences between species. It also refers to differences within species such as differences in age, physical condition (e.g., possible damage to the brain), and genetic makeup (e.g., the presence of certain genes or chromosomes).
4. The nature of the broader context. The examination of how regularities influence behavior will always take place within a certain context. This context always includes specific stimuli (e.g., the room in which an experiment takes place), but may also include other regularities that are also present in that context at the same moment of time. For example, one can study the effect of noncontingent stimulus presentations while the organism is being reinforced for certain actions (e.g., the performance in a burdensome secondary task such as counting back from 1,000 to 1 in steps of three). The point here is that the wider context in which the organism is situated can also moderate the impact of regularities in the environment on behavior.
5. The nature of the regularity. As we previously mentioned, there are different types of regularities in the environment (e.g., regularities in the presence of one stimulus, multiple stimuli, behavior and stimuli, or multiple regularities). Yet it is important to realize that there are also differences within each type of regularity. Each regularity has different facets and each of these facets can be manipulated separately from one another (De Houwer, 2009). For instance, when it comes to noncontingent stimulus presentations, the number, duration, interval between, and the intensity of stimulus presentations can be manipulated. Likewise, the relation between stimuli can be manipulated in many ways, including the number of times the stimuli do or do not occur together, or the temporal and spatial ways in which they are presented. The same goes for the regularity involving a behavior and a stimulus: we can vary the number of times that the behavior is or is not followed by the stimulus, and whether the behavior and stimulus co-occur in different stages of the experiment.

Box 0.1 Genetic Learning

The majority of work in learning psychology has tended to study learning at the level of the entire organism (e.g., by examining whether and when the behavior of human and nonhuman organisms is influenced by regularities in the environment; see Roche & Barnes, 1997). Yet it is also possible to examine the impact of regularities on specific parts of an organism. Doing so requires only that we can establish that the behavior of specific parts of the organism (e.g., the brain, optical system) changes due to events in the environment. This “micro-” or “suborganismic” perspective offers interesting new possibilities for learning psychology. One such possibility is genetic learning (i.e., changes in the activity of genetic material as the result of regularities in the environment). Research has shown that our genetic material responds to certain events in the environment. For instance, certain parts of our genetic material will become active when confronted with a stressful event. More recently, it has been established that the activity of the genetic material can change during the life of an organism as a result of certain experiences (see Bjorklund, 2018; González-Pardo & Álvarez, 2013; Masterpasqua, 2009). For example, Weaver et al. (2004) showed that repeated licking and grooming of baby rats by their mother (a regularity in the environment of the baby rats) results in less activation of specific genes in the hippocampus when those rats are confronted with a stressful event as adults (a change in the behavior of those genes). So, one could say that regularities in the environment lead to changes in the behavior of genetic material and thus that genetic material can learn. Once we have taken this perspective, we can start researching the conditions under which genetic learning occurs. In this way, we would be better able to predict when changes occur in the activity of genetic material and how we can control such changes. In other words, we can try to apply all our knowledge about learning in general to the study of genetic learning in particular.

Most experimental studies in learning research include manipulations of at least one potential moderator of learning. In this book, we will organize learning research on the basis of the sort of moderator that is being manipulated.

0.3.1.2   Abstract functional knowledge   By conducting studies in which potential moderators of learning are manipulated, researchers generate functional knowledge about learning—that is, knowledge about the environmental factors of which learning is a function. An important point to appreciate is that functional psychologists are not satisfied with only accumulating functional knowledge about individual or specific moderators. Rather, they want to create more abstract types of functional knowledge that can be used to predict and influence a wide variety of behaviors across a wide variety of situations (i.e., they engage in the exercise of abstraction). Although it might be tempting to view abstraction as referring to an increase in the complexity of our analyses, this would be a mistake. If anything, abstraction refers to the act of simplifying: certain properties of the situation, organism, and context are disregarded and the researcher’s focus is centered on one or a limited number of properties that apply across a wide variety of cases. Saying that researchers aim to develop abstract functional knowledge thus means that they attempt to distill out those core aspects of the relation between (regularities in) the environment and behavior that apply across many different situations.

To illustrate, consider the well-known studies by Pavlov (1927) on classical conditioning in dogs (see Todes, 2014, for a more extensive and historically accurate description). In a typical experiment, Pavlov repeatedly gave food to his dogs, and just before he administered that food, he rang a bell. In doing so, he created a regularity in the presence of two stimuli: the bell and the food. After several bell-food repetitions, he set out to determine if there was a change in the dogs’ behavior (salivation) when they heard the bell by itself. He indeed found that their salivation systematically increased when they heard the bell, as a function of the number of times that the bell was previously paired with the food. It might be tempting to view this study as being about only salivation and food and to think that the findings might be of interest only to a food expert or physiologist. In fact, Pavlov himself initiated this research based on his interest in digestion and in 1904 was awarded the Nobel Prize for physiology and medicine for this work. But for a learning psychologist, the changes in salivation become really interesting only if we abstract (discard) away from the topographic (superficial) characteristics of the specific stimuli (bell or food) and reactions involved (salivation) and search for the factors that are likely to apply across many different stimuli, behaviors, organisms, and contexts. For instance, we can conceptualize the bell as a conditional stimulus (CS), the food as an unconditional stimulus (US), and the increased salivation due to the bell-food pairings as a conditional response (CR).

When viewed through this lens, we can see that Pavlov’s study with dogs, food, and bells is just one example of an abstract functional principle: presenting a CS and US together can lead to a CR. What is really remarkable is that this general principle of learning (classical conditioning) can be applied to all kinds of stimuli, behaviors, and organisms in all kinds of contexts. For instance, imagine that your friend receives a nasty bite from a dog and subsequently develops a fear of dogs. This can also be seen as one example of the general principle of classical conditioning: the dog is the CS, the bite received from the dog is the US, and the increase in fear of dogs is the CR. The key point here is that the broad applicability of functional knowledge is possible only if we can generate abstract concepts such as CS, US, and CR that allow us to describe a situation without referring to superficial characteristics.

The very same type of abstraction yielded the principle of operant conditioning. Imagine that a rat receives a tasty piece of food each time it presses a lever, and that the rat tends to press the lever more in situations where doing so leads to food. Strictly speaking, such a study is only about the influence of food on lever pressing in rats. But if we once again take a step back and abstract away all the superficial characteristics of the behavior (pressing on a handle) and the stimuli (the food chunks), we will see that this specific (lever-food) effect is just one instance of the more general functional principle of operant conditioning (i.e., behavior is influenced by its consequences). In fact, this example involves just one type of operant conditioning known as reinforcement, wherein a response increases in frequency due to its consequences. Once again, the remarkable feature of abstract principles of learning (like operant conditioning) is that they can be applied to a vast array of stimuli, behaviors, organisms, and contexts (not just rats pressing levers for food). For example, they allow us to understand why people behave in all kinds of ways, such as their tendency to put money into a vending machine (because it leads to a can of drink), compulsively check their social media accounts (because it leads to new information or validation), study for exams (because it leads to good grades), or study the psychology of learning (because it leads to a better understanding of why humans think, feel, and act in the ways that they do). In other words, the abstract principle of operant conditioning tells us that behavior typically depends on its consequences.

So, one of the main goals for functional learning psychologists is to generate abstract principles (such as classical or operant conditioning) in order to account for specific as well as general classes of behavior. It is important to appreciate that the principles of learning are concerned only with the function that stimuli and behaviors have in a given context (i.e., the way in which they are related to other elements in the environment). In classical conditioning, for example, one examines whether a stimulus (CS) elicits a response (CR) following its pairing with another stimulus (US). Whether a stimulus is considered to be a CS or a US does not depend on the superficial characteristics of a stimulus (i.e., whether it is a bell or a buzzer or a dog, or food biscuits or food chunks, or a painful bite of a dog). Rather, we can determine this only by examining the role (or function) that the stimulus has in a given situation (i.e., a stimulus functions as a CS if the reactions to that stimulus change as the result of stimulus pairings; see chapter 2). The very same is true in operant conditioning and reinforcement. Consequences and behaviors are defined in terms of their functions in a given situation. For instance, one would label a consequence a “reinforcer” whenever it leads to an increase in the frequency of a behavior. So in our example of the rat pressing the lever for food, we could describe the food as a reinforcer (i.e., it reinforces the probability of lever pressing). But many different types of stimuli could function in the very same way, whether they involve the delivery of water to a dehydrated rat, warmth to a freezing rat, or access to a running wheel for an exercise-deprived rat. The takeaway message is that for most functional learning researchers, the thing that counts is the function of a stimulus or behavior: what is its role within the relation between environment and behavior.

The approach that focuses on the function of stimuli is also called the analytic-abstractive functional approach (see Hayes & Brownstein, 1986; Hughes, De Houwer, & Perugini, 2016).⁷ It is worth noting that the overarching goal of this approach (i.e., to develop abstract knowledge or principles that explain many different behaviors) does not come at the cost of explaining individual behaviors. In much the same way that the boiling point of water depends on the local air pressure (and can thus be different on Earth than on the moon), the success of general principles in explaining behavior depends on specific environmental factors. Put another way, the results of individual experiments in the psychology of learning are useful not only for formulating general principles of learning (i.e., demonstrating when certain moderators have no influence on behavior) but also for contextualizing those principles (i.e., demonstrating when those moderators do have an influence on behavior).

Both are equally important. It is good to know that classical conditioning occurs in almost all animal species but also useful to know that there are important differences in the conditions under which different animal species show conditioning. The same goes for operant conditioning and other forms of learning. The point here is that there is a two-pronged approach to research centered on abstract or general principles of learning: in some cases, the goal is to formulate such principles by abstracting from individual studies, and in other cases, the goal is to contextualize those principles and show when and how they apply to a given situation. Thus, in the functional approach, researchers recognize the importance of the individual but strive for general knowledge through abstraction on the basis of function.

0.3.1.3   Why strive for abstract functional knowledge?   Although there are many reasons why researchers might adopt a functional approach to the psychology of learning, an important one is that doing so allows them to influence behavior (Hayes & Brownstein, 1986).⁸ The ability to influence behavior not only has practical implications (e.g., for helping people change unwanted behaviors, as in the case of addiction) but is also a vital aspect of explaining behavior. In order to influence a given behavior, one has to intervene on variables that causally produce that behavior. Hence, influencing a behavior implies the ability to explain the behavior in terms of its environmental causes (e.g., regularities in the environment). Explanations of behavior in terms of environmental causes are typically referred to as functional explanations of behavior. Functional knowledge about learning (i.e., knowledge about which environmental regularities influence behavior under which conditions) thus provides functional explanations of behavior.

To illustrate, let us return to our example of the rat that receives a tasty piece of food every time it presses a lever. Based on our knowledge of operant conditioning (and more specifically, reinforcement), we could influence in specific ways how the rat will act in the future. That is, we could say with relative certainty that it will press the lever more often whenever doing so is followed by an appetitive consequence such as food. We can also influence how the rat will behave by manipulating the specific type of relation between behavior (lever pressing) and its consequences (food; see chapter 3). Functional knowledge about learning can also help us influence human behavior. For instance, imagine a situation in which posts on a social media site (e.g., Facebook) are socially reinforced by receiving “likes.” Facebook knows that users will post messages more often if this leads to appetitive consequences such as “likes.” It is therefore a public secret that online companies such as Facebook are only too happy to use knowledge from the psychology of learning to influence the behavior of its users. Such influence will only increase as people become more active in the virtual online world.

Although the general principles of learning can be used for personal or selfish reasons, they can also be used for more prosocial or commendable purposes. For example, many forms of psychotherapy are based on the idea that psychological suffering (e.g., unbearable fear) stems from, and is maintained by, regularities in the environment. Pathological fear of dogs can be the result of previous experiences such as being bitten by a dog. Pathological hand washing can be maintained by the feeling of relief that follows from washing one’s hands. Psychotherapy is therefore often set up to bring people into contact with new regularities, or to modify existing ones (e.g., exposure to dogs without negative consequences). This makes (functional) learning psychology one of the most applicable disciplines in psychological science: it produces knowledge about causes of behavior (regularities in the environment) that are directly observable and that are (often) directly manipulable. Throughout this textbook we will pay attention to the many applications of learning psychology and devote an entire chapter to what we call applied learning psychology (chapter 5).

0.3.2   The Cognitive Approach within the Psychology of Learning

0.3.2.1   Mental mechanisms as mediators of learning   As we mentioned previously, there are two different ways that people usually approach the study of learning. In the previous section, we described the functional approach, which sets out to functionally explain behavior in terms of the environment. The second approach, known as the cognitive approach, involves a very different goal. Whereas functional learning researchers aim to explain behavior in terms of regularities in the environment, cognitive learning researchers want to explain learning (i.e., the impact of environment on behavior) in terms of mental mechanisms. Cognitive learning researchers assume that regularities in the environment can have an impact on behavior (i.e., that learning can occur) only because those regularities set in motion mental processes that produce mental representations that in turn cause the change in behavior. This mental mechanism therefore mediates learning: it is a necessary causal step between the environmental regularity and the change in behavior. Note that we use the word cognitive as a synonym of mental, where mental refers to units of information (mental representations) and the processing of information (mental processes; Neisser, 1967). The key message here is that functional learning psychologists search for moderators of learning, whereas cognitive learning psychologists search for mediators of learning.

But what exactly is a mental mechanism? Well, mental mechanisms are metaphorically similar to physical mechanisms (Bechtel, 2008). In both cases, the mechanism involves a sequence of states through which a certain input leads to a certain output. Consider, for example, a car. In a car, there is a physical mechanism whereby turning the ignition key (input) eventually leads to the car’s propulsion (output): turning the ignition key leads to the ignition of petrol in an engine, which causes cylinders to be moved, and this leads to the turning of the wheels. The mechanism is thus a sequence of steps in which each link acts on the next link, a bit like how one cog in a machine can act on the next cog. Mental mechanisms are very similar to physical mechanisms except that the parts that act on each other are informational rather than physical. In other words, mental mechanisms are collections of mental representations in which the information contained in these representations is processed step-by-step (Bechtel, 2008). To illustrate these ideas, let’s return to the example of Pavlov’s dog. Whereas functional researchers would attribute the change in behavior (salivation) to the prior pairing of a bell and food, cognitive researchers want to (a) explain why pairing a bell and food leads to a change in salivation, and (b) do so by searching for some mental mechanism that might mediate the relation between the environment and behavior. There are many possible mechanistic explanations for this environment-behavior relation. One is that the pairing of the bell and food leads to the formation of associations between the representation of the bell and representation of the food in memory. More specifically, each time the bell and food are presented in memory, their representations in memory are co-activated, which results in a gradual strengthening of the association between the two representations (see Hebb, 1949, for a neural analogue of this idea). Once the association is strong enough, the presentation of the bell on its own will result not only in the activation of the representation of the bell but also, via the newly formed associations, in the activation of the representation of the food, which in turn results in salivation.

Figure 0.4

Schematic representation of the functional and cognitive approaches to the study of learning.

A strength of cognitive theories is that they also deal with abstract types of knowledge. For instance, mechanisms such as the formation and activation of associations between mental representations apply to all kinds of situations, regardless of the stimuli or behaviors involved. Note, however, that this knowledge (about mental mechanisms that mediate learning) is of a different kind than the knowledge acquired by functional learning researchers (who focus on abstract knowledge about the environmental moderators of learning).

0.3.2.2   Why strive for cognitive knowledge?   So why exactly do certain researchers adopt a cognitive approach to learning research? Or put another way, why do they have the goal to explain learning in terms of mental mechanisms? Two reasons stand out. First, cognitive researchers are often driven by the urge to understand the “underlying substance” of a phenomenon. Especially in the Western world there is a tendency to view a phenomenon as being “really” understood only when the underlying mechanism is known. Learning psychologists who share this tendency will never be satisfied with purely (even abstract) functional knowledge because, for them, this provides only a description of the relation between environmental regularities and observed changes in behavior. A cognitive learning psychologist also wants to understand how regularities lead to changes in behavior. The difference between cognitive and functional learning psychologists therefore has much to do with their philosophical assumptions, scientific goals, and what they ultimately consider a satisfactory explanation to be (Bechtel, 2008; Hayes & Brownstein, 1986).

Second, cognitive researchers also tend to believe that knowledge of underlying mental mechanisms is useful because it leads to new questions that increase the chances that we can predict and influence behavior. Understanding a mechanism requires that we not only observe (to predict) or change (to influence) the input but also know all of the different steps in the mechanistic chain. To illustrate, think back to the example of the car. If we know that turning the key leads to the ignition of fuel in the engine, we can predict the movement of the car not only on the basis of the position of the key, but also on the basis of the amount of fuel present in the fuel tank. We can also use our knowledge of the physical mechanism to make new predictions about driving the car (e.g., that the car will not run when the engine is disconnected from the wheels). Even Skinner, a key proponent of the functional approach, recognized that knowledge about mental mechanisms can provide important added value in predicting and influencing behavior (Skinner, 1953, p. 34).

Although there are good reasons why cognitive psychologists are attracted to (mental) mechanistic explanations, there are important limitations to explanations in terms of such mechanisms. Unlike physical mechanisms (such as the car key and engine), mental mechanisms cannot be observed directly. As Wiener (1961, p. 132) said, “Information is information, not matter or energy.” Consequently, the presence or absence of mental processes and contents can only ever be inferred on the basis of behavior (irrespective of whether that behavior is controlled, involuntary, verbal, or neural). This presents an unavoidable problem to the researcher. The problem is that in order to make such an inference, you already need to know how mental processes and representations influence behavior. If you are not 100 percent sure what the mental causes of a particular behavior are, then you cannot be certain what that behavior says about the presence of certain mental processes and contents. Yet, in order to find out what the mental causes of behavior are, you have to observe behavior that you know is 100 percent due to the assumed mental causes. This often leads to a catch-22 where you are never really sure whether and how you can study certain mental processes and contents (see De Houwer, 2011b; Hughes et al., 2016). Although we can certainly gain insight into the mental mechanisms that are assumed to mediate between environment and behavior (Bechtel, 2008; De Houwer & Moors, 2015), the question arises whether the search for mental mechanisms is a good way to achieve prediction and influence over behavior. Debate continues, but one thing is certain: when you conduct cognitive research with the aim of improving your ability to predict or influence behavior, it is always good to check on a regular basis whether the cognitive theories you are working on actually bring you closer to this goal of prediction or influence (De Houwer, Hughes, & Barnes-Holmes, 2017).

Box 0.2 Latent Learning

The distinction between the functional and cognitive approaches can be nicely illustrated by a phenomenon known as latent learning (Chiesa, 1992; De Houwer et al., 2013; see Jensen, 2006, for a detailed treatment from the perspective of functional psychology). Tolman and Honzik (1930) placed a rat in a maze at Time 1. They allowed the rat to explore the maze and then later returned it to the starting position. The researchers did so a number of times and observed little change in the behavior of the rat. Later on (Time 2), the rat was brought back to the same maze. The researchers now placed food at a certain place in that maze. After the rat found the food the first time, it was once again returned to the starting position. Immediately thereafter, the rat returned to the place where it had previously found the food via the shortest route. Other rats that had not been given the opportunity to explore the maze at Time 1 took much longer to find the food the second time. This work indicates that the rat had indeed acquired knowledge about the structure of the maze at Time 1 despite the fact that (a) its behavior did not change at Time 1 and (b) no reinforcers were present in the maze at Time 1. The knowledge that the rat had acquired at Time 1 therefore remained latent (invisible) until it could use this knowledge to find the food at Time 2.

On the one hand, the phenomenon of latent learning is nothing special for a functional learning psychologist. The change in behavior at Time 2 is, after all, a function of the regularities that the rat experienced when it explored the maze at Time 1. The only unique thing about latent learning is that there is a period of time between experiencing the regularities (experience with the layout of the maze at Time 1) and the observed change in behavior (quickly finding the food again at Time 2). This, however, does not fundamentally change the fact that the change in behavior is due to regularities in the environment. Hence, from a functional perspective, the term latent does not explain anything. It does not refer to some hidden or mental level that mediates between environment and behavior. It is merely a descriptive label that orientates the researcher toward the fact that there was a temporal gap between the occurrence of a regularity and the observation of a change in behavior.

On the other hand, the phenomenon of latent learning is very important and special for cognitive psychologists. These researchers are searching for the mechanism via which the environment can influence behavior. The fact that an experience at Time 1 can have an influence on a later Time 2 proves to them that there must be a mechanism by which the experience at Time 1 can be preserved in some way so that at Time 2 it can lead to a change in behavior. After all, from a mechanistic viewpoint, there must be an immediate cause behind every change in behavior that triggers this change, just as one cog must be set in motion by another cog in a machine (Chiesa, 1992, 1994). In latent learning, the regularities that are present at Time 1 cannot be the immediate cause because the change in behavior takes place when these regularities are no longer present. So there must be a mental representation that is formed at Time 1, which then remains in memory, and it is this representation at Time 2 that is the immediate cause of the change in behavior. This mental representation remains latent (i.e., hidden) until it influences behavior at Time 2. Hence, for cognitive learning psychologists, latent learning offers proof of the existence of mental representations.⁹

Note that latent learning is often used as an argument for the conclusion that a cognitive approach to learning is superior to a functional approach: it shows that one has to assume mental representations in order to explain how regularities in the environment can influence behavior. However, this conclusion is wrong because it loses sight of the fact that cognitive and functional learning psychologists have fundamentally different scientific objectives. As we previously outlined, functional psychologists are looking not for immediate (mechanistic) causes of behavior but rather for functional causes (i.e., environmental regularities that drive changes in behavior). For them, it is enough to know that a behavior is a function of a certain regularity in the environment, because this information allows behavior to be predicted and influenced. Latent learning is crucial only for learning psychologists who have the goal of explaining how regularities in the environment lead to changes in behavior. For them, latent learning offers proof that mental representations must be part of the mediating mechanism.

One final point. According to our definition, learning cannot occur without a change in behavior (regardless of whether this change occurs at the motor, physiological, or neural level; see our definition of behavior). Hence, in the experiment of Tolman and Honzik (1930), the conclusion that the rat has learned makes sense only after the experimenter observes that the rat took the shortest route at Time 2. Once this change in behavior has been observed at Time 2, and provided that it can attributed to the regularities at Time 1, it is justifiable to conclude that learning has taken place.

We realize that our perspective on this issue is unusual. For many, learning is defined as the storage of knowledge rather than as a change in behavior due to regularities. From such a perspective, it is self-evident that learning can occur without a change in behavior so long as knowledge has been stored. Thus, one can define learning in different ways. As we mentioned before, we adopt a pragmatic approach to definitions, one that is less interested in whether a definition is “true” or “correct” in some absolute sense and more interested in whether a definition helps us to better predict and influence behavior. We believe that, in general, the definition we offer throughout this book is useful in this latter sense. Also note that if there is no impact of regularities on behavior, then there is nothing to explain at the cognitive level. Hence, even if one were to define learning as the storage of knowledge, it can be studied only by examining learning as a behavioral phenomenon (see also our reflections on “Think It Through 0.1”).

0.3.3   The Relation between the Functional and Cognitive Approaches in Learning Psychology

The relation between the functional and cognitive approaches is currently less than ideal. Both traditions have their own scientific associations, conferences, journals, and textbooks, and supporters of one approach have little or no contact with supporters of the other. Although there are historical reasons why the relation between the two approaches is so problematic (see box 0.3), there is in principle no reason why this must be the case. Indeed, from the perspective of our functional-cognitive framework, these two approaches are not competing for scientific legitimacy but are simply playing two different scientific games (De Houwer, 2011b; Hughes et al., 2016). On the one hand, the functional approach in learning research wants to explain behavior by first examining which environmental regularities influence behavior under which environmental conditions and then abstracting out general principles that can account for classes of behavior that differ across time and situations. On the other hand, the cognitive approach wants to explain learning (the impact of regularities on behavior) in terms of mental mechanisms. The two approaches are thus situated at different levels of explanation, each with their own explanandum (i.e., the concept that has to be explained) and explanans (i.e., the concept used to explain; see table 0.1).¹⁰

Box 0.3 Behaviorism and the Myth of the Cognitive Revolution

The functional approach has close ties with a tradition in psychology known as behaviorism. Importantly, however, there have historically been several branches of behaviorism. Perhaps the most well-known is methodological behaviorism, as proposed by John B. Watson (1913). Watson’s perspective was very close to logical positivism, a movement within the philosophy of science that stated that “true” knowledge can be obtained only on the basis of objective observation. This led Watson to question the scientific validity of the introspective method, which was dominant in psychology at that time. Introspection requires that people report their subjective sensations and feelings. Watson argued that introspection can never be part of a scientific psychology given that the accuracy of these introspections can never be objectively determined. He instead proposed that only objectively observable stimuli and behaviors should make up the matter of psychological science. He argued that behavior should be explained only in terms of links between stimuli and responses (“behavioral chains”), in which a stimulus leads to a first reaction, which then elicits a second reaction, which then elicits a third, and so on until the behavior of interest eventually takes place. This perspective fits perfectly within a mechanistic approach to science. Indeed, the only difference with the cognitive approach outlined previously is that Watson did not accept that mental representations are part of the mechanism that leads to behavior. The phenomenon of latent learning (see box 0.2) was therefore very problematic for Watson’s perspective.

What should be clear from this section is that Watson’s version of behaviorism is not part of the contemporary functional approach to psychology. Actually, the functional approach to the psychology of learning that we have described in this handbook is very similar to the radical behaviorism of Skinner (1938, 1953). Unlike Watson, Skinner was not interested in mechanisms but rather in functional relations between the environment and behavior. And unlike Watson, Skinner also accepted that subjective thoughts and feelings can also be studied scientifically, provided that thoughts and feelings are considered to be covert behaviors (i.e., behavior that is observable only to the person emitting that thought or feeling).

Today, functional and cognitive psychologists largely go their separate ways, with theories and findings in one approach rarely informing or driving progress in the other. One of the reasons for this problematic relation is that often no distinction is made between the methodological behaviorism of Watson and the radical behaviorism of Skinner. Cognitive psychologists assume that findings such as latent learning (see box 0.2) prove conclusively that behaviorism (as a whole) is wrong and must be replaced by the cognitive approach. This idea is part of the myth of the cognitive revolution (Watrin & Darwich, 2012), which states that cognitive psychology has replaced behaviorism as the dominant approach in psychology, a bit like when one animal species supplants another during natural evolution. This myth propagates the idea that cognitive psychology and behaviorism are competitors in a contest in which cognitive psychology has emerged superior and behaviorism, like the dinosaur, has gone extinct.

As Watrin and Darwich (2012) rightly point out, many questions can be asked about where this myth comes from and why it persists. First, many of the critiques of behaviorism apply only to methodological behaviorism and not to Skinner’s radical behaviorism (e.g., think back on the phenomenon of latent learning). Likewise, the criticism of Skinner’s work as formulated by Chomsky (1959) is also unjustified in many ways (for reasons why, see MacCorquodale, 1970; Palmer, 2006; Watrin & Darwich, 2012) and applies only to Skinner’s theories and not to other theories in functional psychology (Hayes, Barnes-Holmes, & Roche, 2001). Second, it is a misconception to think that behaviorism is extinct. Contrary to popular belief, many functional psychologists still consider Skinner’s work, as well as more recent work in this area, as a source of inspiration for their own research and theorizing. They unite in associations such as the Association for Behavioral Analysis International (ABAI) and the Association for Contextual Behavioral Science (ACBS), both of which have thousands of members worldwide who are influential in applied and clinical psychology. Finally, theories within functional psychology continue to emerge and evolve, as evidenced by, among other things, the development of relational frame theory (Hayes et al., 2001) as an alternative to Skinner’s own theory (1957) of language and cognition.

Nevertheless, the poor relation between these two approaches in psychology is not just the result of misunderstandings on the part of cognitive psychologists. Functional psychologists such as Skinner (1990) are also partly responsible for the poor relation between these two approaches. Instead of seeing both approaches as complementary, Skinner and other functional psychologists continued to ask questions about the scientific character of cognitive psychology, particularly with regard to the use of concepts that refer to unconscious mental processes such as inhibition and spreading or activation. In doing so, they lost sight of the fact that such concepts are unavoidable when the goal is to describe the mental mechanisms that underlie behavior (such as latent learning, see box 0.2). Thus, the poor relation between functional and cognitive psychology can in large part be reduced to a mutual lack of insight in and respect for the nature and objectives of the other approach.

Table 0.1
The concepts that need to be explained (explanandum) and the concepts used to explain (explanans) in the functional and cognitive approaches to the psychology of learning
		Explanandum (Concept that must be explained)		Explanans (Concept used to explain)
Functional		Behavior (e.g., salivation)		Regularities in the environment (e.g., pairings of bell and food)
Cognitive		Learning (e.g., classical conditioning)		Mental processes (e.g., association formation)

What is clear is that both levels of explanation have their respective merits. Therefore, it is difficult, if not impossible, to decide which approach is the “best” or “most important” (questions about what is “best” are prescientific and are decided on the basis of one’s philosophical assumptions; Hayes & Brownstein, 1986). Instead of pitting the two approaches against each other, we believe a more productive strategy would be one in which people accept that (a) the two approaches have very different scientific objectives, and that (b) the objectives important to one approach can be used to evaluate only the products that emerge from that approach and not others, in much the same way that the rules that make sense in one sport (e.g., soccer) cannot be used to govern the activity of others (e.g., basketball; see Hughes, 2018). In other words, neither approach is strengthened by showing the weakness of the other.

Once one accepts that these two approaches are not in competition, it quickly becomes clear that the functional approach can reinforce progress within the cognitive, and vice versa. Take the functional approach: functional knowledge about learning can provide new insights into the mental mechanisms that are assumed to mediate learning. For instance, one can compare existing cognitive theories and evaluate the extent to which they are able to account for the existing body of functional knowledge. After all, a good mental process theory is a theory that can explain why a certain regularity in the environment has an impact on behavior only under certain conditions (i.e., a good mental process theory has a high heuristic value insofar as it can explain existing functional knowledge). The more functional knowledge we accumulate, the better we are able to evaluate and compare mental process theories with one another. This explains why cognitive learning psychologists also carry out experiments in which they manipulate environmental factors and check whether this has an influence on learning (i.e., because doing so allows them to test their mental process theories; see Hughes et al., 2016, for a more nuanced view). What is important to appreciate here is that cognitive learning psychologists do not have to restrict themselves to accumulating their own functional knowledge by conducting experiments. Rather, they can also use the wealth of experimental data already collected by functional learning psychologists. But functional learning psychologists have more to offer than just their data. As we noted earlier, functional psychologists strive to formulate abstract principles on the basis of individual findings. The abstract concepts that they develop when formulating these principles can also be useful for cognitive psychologists. This is because cognitive psychologists tend to describe their data either in very superficial terms or in terms of their cognitive theories, which has all sorts of disadvantages that we will not discuss here. Describing data in terms of abstract functional concepts overcomes these disadvantages and can thus be useful for cognitive psychologists (see De Houwer, 2011b; De Houwer et al., 2017; Hughes et al., 2016).

Cognitive learning research can also help functional psychologists to discover the moderators of learning. This is because a good mental process theory has, in addition to a high heuristic value, a high predictive value (i.e., the ability to make novel predictions about the conditions under which learning occurs). Therefore, testing predictions derived from mental process theories leads to new functional knowledge about the conditions under which learning occurs. This functional knowledge can be used by functionally oriented researchers to refine their own analytic-abstractive concepts, theories, and procedures (see Barnes-Holmes & Hussey, 2016, for a discussion about the possibilities and limitations of what functional psychologists can learn from cognitive psychologists).

The key point here is that there is a natural interaction between observation, functional knowledge about learning (knowledge about the moderators of the impact of regularities in the environment on behavior), and mental process theories about learning (hypotheses about the mental mechanisms responsible for the impact of environmental regularities on behavior), and this interaction can be useful for functional and cognitive learning psychologists alike. Both conduct experiments in which they manipulate environmental factors and observe behavior. Provided there are sufficient controls, they both can arrive at functional knowledge. Because functional knowledge provides insight into the causes of behavior, one can use this knowledge to predict new observations (which behavior will occur under which conditions), which can in turn lead to new functional knowledge. Thus, an interaction between the level of observation and functional knowledge is possible.

But there can also be an interaction between the functional and cognitive level of explanation. Based on existing functional knowledge, cognitive psychologists can try to design theories about the mental causes of learning. These theories can in turn lead to new predictions about functional relations (i.e., about moderators of learning), which are tested in new experiments, leading to observations that in turn can result in new functional knowledge. This interaction between observation, functional knowledge, and mental process theories is depicted in figure 0.5 (for more on this idea, see box 0.4).

Figure 0.5

The interaction between observation, functional knowledge, and theories about mental processes.

Box 0.4 Distinguishing Two Levels of Functional Knowledge

If researchers stick to the results of individual experiments, they will be able to say something only about the specific effects that occur in those experiments—that is, the impact of those specific aspects of the environment (e.g., the pairing of bell and food) on those specific types of behavior (e.g., salivation). The functional knowledge that is generated by individual experiments has some merit (i.e., it allows for prediction and influence within that particular setting), but its merit is restricted to the effect that is being studied. Hughes et al. (2016) therefore referred to the functional level of the individual experiment as the effect-centric functional level. Scientists, however, also want to use the data of individual experiments to arrive at more abstract knowledge that can be applied to a wider range of situations, including situations never encountered before. As noted earlier, functional psychologists achieve this by formulating abstract principles that refer to the function of events. For instance, the principle of reinforcement states that behavior increases in frequency if it is followed by a reinforcer; this principle can be applied not only to rats pressing a lever for food but also to any other organism that emits behaviors that have consequences. This type of knowledge is functional in nature (it is about environment-behavior relations) but it is formulated in abstract functional terms rather than in terms that refer to the stimuli and behavior involved in a specific experimental setup. Hughes et al. (2016) referred to this level of functional knowledge as the analytic-abstractive functional level. Developments at the analytic-abstractive functional level are driven by developments at the effect-centric functional level. Likewise, developments at the analytic-abstractive functional level give rise to new predictions that produce new effect-centric functional knowledge. In sum, the two functional levels interact in mutually supportive ways.

Cognitive psychologists also aim to develop abstract knowledge that can be applied in a wide range of situations. They also formulate this knowledge on the basis of individual experiments and thus also draw on effect-centric functional knowledge. However, cognitive psychologists typically formulate their abstract knowledge in terms of mental constructs (e.g., association formation, inhibition, working memory) that (ideally) can be applied across a range of situations.

Figure 0.6 provides a schematic representation of this extended analysis of levels of explanation in psychological research. The increase in complexity of this analysis is counterbalanced by a more fine-grained view on ways in which functional and cognitive researchers can interact. First, both approaches can feed into the effect-centric functional level and thus into developments at both the cognitive level and the analytic-abstractive functional level. Second, cognitive psychologists can benefit from knowledge and concepts developed at the analytic-abstractive functional level, just like functional psychologists can find inspiration in cognitive theories when formulating abstract functional principles.

Figure 0.6

An extended analysis of levels of explanation in psychology.

Critically, this interaction between the two approaches can succeed only if they are clearly distinguished from each other. Take the interaction between functional knowledge and mental process theories. Separating these two levels implies that elements of functional knowledge (i.e., behavioral effects) are not treated as interchangeable with elements of mental process theories. Mixing the functional and mental levels occurs whenever one starts to define behavioral effects in terms of mental processes. Unfortunately, this tendency is all too common in the psychology of learning and elsewhere in psychological science. To illustrate, take classical conditioning. Many researchers mistakenly believe that classical conditioning effects are synonymous with the formation of associations between mental representations in memory. Conflating the thing that needs to be explained (effect) with the thing that is used to explain (mental process) has many disadvantages (see De Houwer, 2007, 2011b; De Houwer et al., 2017).

One major disadvantage is that it quickly becomes difficult to actually study classical conditioning. After all, from this perspective, it is not enough to establish that a regularity in the presence of multiple stimuli leads to a change in behavior. Researchers also have to show that the impact of stimulus pairings is mediated by the formation of associations in memory. Yet, as we noted earlier, it is very difficult to establish that mental representations are present (especially given the nonobservable nature of information; Wiener, 1961). This makes it very hard to establish that classical conditioning has actually occurred, and thus difficult to study the phenomenon. A second disadvantage is that theories about mental causes often change. For example, new theories have emerged that question the very idea that learning is due to the association formation and activation (Mitchell, De Houwer, & Lovibond, 2009). Now if one were to define classical conditioning mentally (in terms of association formation) and to decide that associations do not actually exist, then one should conclude that classical conditioning does not exist either.

Critically, however, the fact that a theory about a phenomenon is wrong does not necessarily mean that the phenomenon itself does not exist. Quite the opposite: classical conditioning as an effect (i.e., the impact of stimulus pairings on behavior) exists irrespective of whether the mental explanation of that effect (in terms of association formation) is correct or not (Eelen, 1980/2018). It is therefore essential that a strict distinction be made between what the mental process theory sets out to explain (e.g., classical conditioning effects) and elements of the mental process theory itself (e.g., the formation of associations in memory). In the case of learning, this means a distinction must be made between the impact of environmental regularities on behavior (i.e., learning as an effect) and the mental processes that are assumed to mediate between these regularities and behavior (i.e., learning as a mental process). In combination with the distinction between procedures and effects that we discussed earlier (see section 0.2.3), we can therefore conclude that we define learning (both in general and the different types of learning) as an effect and not as a procedure or a mental process. It is this distinction between procedure, effect, and mental process that allows us to create an interaction between the functional and cognitive approaches to learning.

0.4   Structure of the Book

We have organized our book in the following manner. The first three chapters are devoted to the three forms of learning that have traditionally been studied in this field: the effects of noncontingent stimulus presentations (chapter 1), classical conditioning (chapter 2), and operant conditioning (chapter 3). In chapter 4 we discuss more complex forms of learning in which multiple regularities are involved. Each of these chapters begins with an overview of the most important functional knowledge we currently have on each type of learning—that is, the factors that are known to moderate the type of learning that is addressed in that chapter (e.g., the nature of stimuli and/or behaviors, the nature of the observed behavior)—and then we turn our attention to the mental process theories that currently dominate our thinking on each type of learning. Finally, chapter 5 provides an overview of how the psychology of learning has contributed to solving real-world problems such as psychological suffering.

Think It Through 0.3: What Is the Relation between Functional and Mental Process Explanations?

Which of the following statements is correct?

• If you have a functional explanation, then by definition you also have a mental process explanation.
• If you have a mental process explanation, then by definition you also have a functional explanation.

Think It Through 0.4: What Is the Relation between Cognitive and Neural Explanations of Learning?

The cognitive approach offers an explanation for learning in terms of mental processes and content. However, learning can also be explained on the basis of other processes such as neural processes in the brain. For example, it can be said that learning is the result of the formation of new connections in the brain.

What do you think? What are the differences and similarities between a statement of learning in terms of the brain and a statement of learning in terms of mental processes?

The book you are now reading differs considerably from other textbooks on the psychology of learning, especially in the way it organizes knowledge on learning. For instance, in many cognitively inspired textbooks, no clear distinction is made between functional knowledge and mental process theories (e.g., Bouton, 2007, 2016; Domjan, 2000; Schwartz, Wasserman, & Robbins, 2002), which often leads readers to confuse learning phenomena (e.g., classical conditioning) and mental process theories (e.g., association formation). Likewise, in textbooks written by functional psychologists, little or no attention is paid to the mental process theories of learning (e.g., Catania, 2013; Michael, 2004; Pierce & Cheney, 2008). This is regrettable, given the heuristic and predictive value of those theories. The functional-cognitive framework that lies at the heart of this book celebrates and embraces both approaches to learning psychology. This book is unique because it combines insights from both approaches.

Another important difference between this and other books on learning research is that we explicitly limit ourselves to the psychology of learning. The titles of many other books contain the phrase “learning and behavior.” They discuss not only learning (changes in behavior that are caused by regularities in the environment) but also other causes of behavior (e.g., a single stimulus at one point in time, genetic factors). As we noted at the start of this chapter, we believe that behavior that is a function of a single stimulus (e.g., fear response to a loud bang) does not qualify as learning. A discussion of genetic factors does fit in a book on the psychology of learning insofar as it concerns the moderating impact of genetic factors on learning (see Think It Through 0.2). However, behavior that is determined only by genetic factors and not by regularities in the environment does not qualify as learned behavior, and thus falls outside the scope of a book on learning.

Before we delve into the first chapter, we should make several things clear. First, we cannot discuss every study or insight relevant to learning. That would be impossible, given that the discipline has existed for more than one hundred years and was for a long time the dominant topic of research in all of psychological science. A wealth of theories, procedures, and findings have accumulated over the decades, and we can cover only a thin slice of that knowledge in this book. For this reason, we focus on those insights that have had a major impact on our understanding of learning and, where possible, we consider potential applications of these findings for domains elsewhere in psychology and society. Nevertheless, we hope to provide a representative and cutting-edge picture of this research area.

Second, we pay little or no attention to the growing literature on the role of neural processes and states within learning. This research fits within the so-called behavioral and cognitive neurosciences. Although neural processes and states certainly fit within the functional-cognitive framework outlined in this introductory chapter (see box 0.5), it would require considerable space to provide even a limited overview of this research area. For readers interested in acquiring information on behavioral and cognitive neuroscience, we recommend Breedlove and Watson (2016) and Gluck, Mercado, and Myers (2016). Third, connectionist models also will not be described. These models can be regarded as simulated neuronal systems and thus as belonging to neuroscience (see Clark, 1990; De Houwer, 2009).

Box 0.5 What Is the Role of the Brain in Learning Research?

Neural structures (parts of the brain) and processes (brain activity) can be involved in the study of learning in different ways. In contrast to mental processes and contents, neuronal structures and processes are observable and directly manipulable because they are part of the physical universe. Therefore, these structures and processes can be included in a functional approach to learning research, either as an independent or dependent variable (see also Vahey & Whelan, 2016). For instance, one can treat neuronal structures and processes as independent variables that are manipulated (e.g., through brain surgery or administering chemical substances that influence the activity of the brain) and examine how doing so impacts learning (i.e., the impact of environmental regularities on behavior). In this way, neuronal structures and processes can be seen as elements of the environment (in a broad sense, the organism is also part of the environment) that potentially moderate learning. One can also treat neuronal processes as a dependent variable (i.e., as a form of behavior) and examine the conditions under which environmental regularities lead to changes in these neuronal processes. This would constitute a functional study of neuronal learning: how does the behavior of the brain change as a result of regularities in the environment? (see also box 0.1 about a comparable study of genetic learning). Both approaches can be found in a research domain called behavioral neuroscience (see Breedlove & Watson, 2016, for an overview).

In addition to a functional approach of the brain in learning research, one can also adopt a neural mechanistic approach. This involves looking for the neural mechanisms that mediate learning. This assumes that neural structures and processes are necessary links in a neural mechanism through which environmental regularities can influence behavior. This approach differs from a functional approach to learning psychology because attention is centered on the search for neural mechanisms instead of functional knowledge (i.e., environment-behavior relations). Certain functional psychologists are also interested in neural mechanisms, given that they help them achieve prediction and influence over behavior. After all, neural mechanisms consist of physical components (such as parts of the brain and chemicals) that are directly observable and manipulable (see also Skinner, 1953, p. 34, who acknowledged the importance of physical mechanisms). But for a functional psychologist, knowledge about mechanisms is at best only a means and never an end in itself; the main goal will remain one’s capacity to influence behavior. Once again, we see that the scientific approach of a researcher is determined by his or her goals, not by what type of research he or she engages in.

Finally, one can also examine the brain from a cognitive approach to the psychology of learning. Both functional knowledge about the role of neural structures and processes in learning as well as knowledge about the neural mechanisms that mediate learning can be used as input for cognitive theories of learning. This becomes possible as soon as we make assumptions about which mental processes are carried out by which neural structures and processes. If we then see, for example, that a certain neural structure is involved in a certain form of learning, and we know that this structure is responsible for storing certain information, then we can decide that storing that information is crucial in that form of learning. It is, however, important to realize that there are risks in making inferences about mental processes on the basis of activity at the neural level. Even if we look at the brain, we can never directly observe mental representations or processes. We can only ever make assumptions about what information is processed in the brain by examining how the brain responds to stimuli in the environment (the brain as a dependent variable) and how interventions in the brain change behavior (the brain as an independent variable). And so long as there is uncertainty about the assumptions underlying the relation between brain and mental mechanism, there will be uncertainty about what we can learn about mental processes based on neural research (see also Poldrack & Yarkoni, 2016). Consequently, it seems easier to evaluate theories about neural rather than mental mechanisms.

We hope that once you finish reading this book you come to see that the psychology of learning is concerned with (and thus can help you better understand) the very “source code” that underpins the behavior of human and nonhuman animals. Equipped with an understanding of the behavioral principles, theories, and findings discussed in this book, you will be better able to examine many behaviors in daily life and ask questions about the regularities that give rise to and sustain them. Those interested in clinical psychology can ask whether arachnophobes are afraid of spiders because they previously had a negative experience with spiders (e.g., a painful bite) or their phobia is based on complex learning (e.g., negative verbal information about spiders). Those interested in marketing research can better identify whether people buy a certain product because earlier purchases were followed by positive consequences (e.g., social reinforcement from others) or merely because the product was paired with other positive stimuli (e.g., a nice advertisement). Readers interested in public policy can better understand the factors that influence pressing problems created by human behavior, such as climate change, conflict, overeating, overpopulation, and resource depletion. Once you can read the source code that drives human behavior, you will be better positioned to predict and influence that behavior yourself. Learning research is an essential part of psychology. It is this vision that we aim to convey throughout this book.

Notes

1. 1. Skinner (1953) divided behavior into two categories (i.e., overt and covert behavior). Overt behavior refers to responses that are observable by third parties, whereas covert behavior refers to responses that can be observed only by the organism itself. One could have long philosophical discussions about whether conscious thoughts, perceptions, and feelings qualify as behavior. Skinner ignored discussions about the “true nature” of behavior and took a purely pragmatic position: it may be useful to “act as if” thoughts, perceptions, and feelings are all types of behavior. After all, acting as if thoughts, perceptions, and feelings are types of behavior entails that the learning principles that apply to behavior in general can be applied to thoughts, perceptions, and feelings as well (for more on this, see the section on relational frame theory in chapter 3).

2. 2. We use the term regularity in the sense of something that is orderly, regardless of whether the orderly pattern is limited to one point in time or repeated across time. Even the co-occurrence of two stimuli at one point in time can be considered orderly and thus a regularity. We realize that the latter statement does not sit well with an alternative definition of regularity as something that is periodic (i.e., occurring at fixed intervals in time (see https://www.thefreedictionary.com/regularity), but we hope that readers are willing to set aside this alternative view in the context of this book.

3. 3. There is no a priori reason why the presentation of two stimuli at a single point in time could not influence behavior at a later point in time. For instance, a single pairing of a tone and shock could increase fear responding toward the tone when it is encountered again at a later point in time. The presentation of one stimulus at two moments in time could also be seen as a regularity that has the potential to influence behavior. For instance, imagine that the reaction evoked by a loud noise is stronger the first time the noise is presented than the second time it is presented. This change in behavior would qualify as an instance of learning if it can be argued that the reduction in the reaction to the noise on its second representation is due to the regularity in the presentation of the noise—that is, to the fact that it is presented for a second time. It can, however, be difficult to separate learning from priming. In a typical priming procedure, on each trial there are two stimuli that occur together in time and space (e.g., the word DOCTOR and the word NURSE). Hence, priming procedures involve a regularity in the environment (e.g., two words that co-occur). There is also a change in behavior (e.g., the time taken to decide that NURSE is an existing word is shorter when NURSE is preceded by the word DOCTOR than when it is preceded by an unrelated word). One could argue, however, that in priming effects, the change in behavior is not caused by a regularity (e.g., the singular co-occurrence of the two words) but by the presence of a prime stimulus (e.g., the word DOCTOR). According to this analysis, priming effects are not instances of learning because the cause of the change in behavior is a single stimulus at one point in time (i.e., the prime), rather than a regularity in the presence of stimuli (i.e., the pairing of the prime and target). The picture gets more blurred when we consider priming studies in which the prime and target are identical (i.e., repetition priming). This also leads to a change in behavior (i.e., responding to the target is faster when preceded by an identical prime), but it is unclear whether this change is due to the presentation of the prime (in which case, the change in behavior would not qualify as an instance of learning) or to the fact that the target stimulus was already presented before (in which case, it would qualify as an instance of learning). One the one hand, problems with distinguishing priming and learning again illustrate that claims about learning are at best hypotheses about the causes of changes in behavior rather than “real” facts to be discovered. On the other hand, one could argue that these problems reveal the limitations of our definition of learning. As we noted at the start of this chapter, definitions are rarely perfect. We also noted that our definition should be regarded as a working definition rather the “true” definition of learning. Its only aim is to facilitate research and application. We hope that the remainder of this book shows that our definition serves this aim well.

4. 4. Some researchers use the term classical conditioning only when one of the two stimuli has a biological relevance (e.g., food, painful stimuli). Although historically such stimuli are often used in research on classical conditioning, we see no reason to limit research on classical conditioning to these kinds of stimuli. Moreover, it is often difficult to determine what counts as a biologically relevant stimulus (De Houwer, 2011b).

5. 5. Likewise, a distinction must also be made between adaptation and adaptive. While adaptation implies only that behavior is influenced by regularities in the environment, adaptive refers to achieving a certain norm or end goal (e.g., reproduction). Adaptation therefore implies only a change in behavior as a result of regularities in the environment, whereas adaptive implies something more (i.e., that there is a certain direction in which the behavior changes).

6. 6. Note that we make a distinction between moderation of learning and moderated learning. As proposed by De Houwer et al. (2013), moderated learning refers to situations in which the effect of regularities on behavior (i.e., learning) depends on another regularity (see section 0.2.2). This therefore concerns situations in which learning is moderated by one specific type of moderator: another regularity in the environment. When learning depends on other types of moderators (e.g., the nature of the stimulus), we do not talk about “moderated learning” but about “moderation of learning.” We adhere to this position but add the requirement that moderated learning should always involve a change in behavior. Hence, we define moderated learning as a change in behavior that is the joint effect of multiple regularities. This definition implies that the change in behavior would not occur in that manner if one of the regularities would be absent. The additional requirement allows us to exclude situations in which a regularity in the environment eliminates learning (e.g., when presenting a stimulus on its own eliminates the impact of stimulus pairings on behavior), which cannot be instances of learning (in the way that we defined learning) because there is no change in behavior.

7. 7. Throughout this handbook we will refer to the “functional approach” as a shorthand for the analytic-abstractive functional approach.

8. 8. Another reason is to predict behavior. For many functional researchers, however, prediction is relevant only to the extent that it can aid influence on behavior (see Hayes & Brownstein, 1986).

9. 9. Although less crucial in the present context, cognitive researchers also see Tolman’s studies as proof of the mediating role of motivational factors. More specifically, for them, it shows that it is not enough to have a mental representation to set behavior in motion; motivation to deploy that representation is also needed (e.g., deploy knowledge about the maze only when it is useful for locating a desired food).

10. 10. Some readers might be reminded of Marr’s (1982) popular distinction between the computational, algorithmic, and implementational levels of analysis. One could, however, argue that Marr’s levels of analysis are all situated at the cognitive level of explanation in that they are directed at understanding the way that mental processes mediate behavioral phenomena. More specifically, computational analyses specify the input and output of a mental process, algorithmic analyses reveal the information processing steps via which the input is transformed to the output, and implementational analyses uncover the way in which mental processes are realized at the physical level. From this perspective, even cognitive researchers who analyze mental processes at the computational level operate at a different level of explanation than functional researchers. Although functional researchers also describe input-output relations, they do so not to improve understanding of mental processes but to better predict and influence behavior (Hayes & Brownstein, 1986). Moreover, functional researchers consider only inputs and outputs that are part of the environment (i.e., mental inputs and outputs are not taken into account; see De Houwer & Moors, 2015).