Chapter summary

An important underlying assumption behind cognitive psychology is that learning and behaviour are governed by laws. As we discussed in Chapter 1, just how such laws might be discovered has been one of the main sources of controversy since the earliest scientific psychological investigations began. Despite the turbulent history of psychology, the methodology and the basic assumptions behind the study of animal learning appear to have been indifferent to these changes. Yet in practice, animal learning research has been at the helm of the controversy. The study of connectionism (Chapter 9) has also been influenced to some degree by animal learning theory, the common link being the doctrine of the associationists who argued that the unit of memory is the association of ideas, and learning theorists have provided some basic rules about how associations can be learned. For these reasons, it is important to consider the main elements of learning theory.

Classical conditioning

In this section you will read about the roots of learning theory and how the study of animals can bring insights into the study of human learning. You will study Pavlov's conditioning procedures and the main phenomena of classical conditioning.

Animal learning

One reason why animal learning theory has been so prominent in shaping the development of psychology is because it is relevant to human learning and human memory. On the relevance to humans Pearce (1987) notes that ‘an accurate model of animal intelligence would provide a tremendous spur to its study in humans', and armed with such knowledge it is likely that ‘considerable insights into the workings of the human brain will follow'. Animals are less complicated than humans: their behavioural repertoire is comparatively much simpler and their brains less developed. Despite the differences, the similarities are such that valid comparisons between human and animal learning and memory may be made.

Some of the neurophysiological similarities are quite astounding: the same types of neurons are found across species, including humans; all neurons communicate via synapses, and with similar substances. In fact neuronal ‘signals are so similar in different animals that even a sophisticated investigator is unable to tell with certainty whether a photographic record of a nerve impulse is derived from the nerve fibre of a whale, mouse, monkey, worm, tarantula, or professor' (Kuffler, 1984). To some extent, much of the data on animal learning can be observed in humans. Some researchers have suggested that basic laws of conditioning appear to be universal across species (e.g. Turrkkan 1989). Animal learning theorists seek to discover how animals learn about the relationships between events in their environment. Some events (such as the arrival of a predator) can be predicted from other events (such as the appearance of a shadow) and animals need to learn about such relationships to survive.

Animal learning theory began as an attempt to formalise ideas from the associationists (David Hume, John Locke, John Stuart Mill, F. C. Bartlett, and S. Lockery). The basic assumption of associationism was that mental phenomena consist of sensory impressions, and internal copies of these are linked together through ideas. They argued that the symbols used in any culture, especially religion, are examples of the formation of associations. For example, the cross is a Christian symbol and is associated with ideas such as sacrifice and devotion.

For the associationist school (e.g. Hume, Hartley, J. S. Mill, Locke), the existence of mental associations is intuitively obvious. For example, a picture of a joint of meat can elicit its smell, the sound of a voice can elicit a face, a particular song may elicit personal memories, and so on. However, the associationists argued that such links were not confined to associations between sensations, but also to other more complex forms, such as conceptual or linguistic relationships. For example, each line of the Lord's Prayer may be a cue for the next line, and each letter of the alphabet may cue the next letter.

The associationists attempted to derive ‘laws' of association. The most fundamental of these were that memory of events, objects, people, ideas and so on are linked through (1) a contiguity in space, and (2) a contiguity in time. Hence events occurring in the same place or at the same time become associated with each other, such that the sensory impression of one may trigger the mental representation of the other. It is interesting to note that cause-and-effect was not considered logically justified. Indeed, Hume stated that as a philosopher of science he did not believe in cause and effect, yet as a human he did since the inference of cause-and-effect relationships is essentially human.

As learning theory has progressed, most of the controversy has surrounded the question of contiguity versus contingency when asking what the precise relationships that animals learn are. Contiguity is a view that we learn a relationship through mere contact in time (we learn that two things are associated because they occur together in time); for example, we ‘learn' that two people are friends because they are always seen together. Contingency refers to learning that something predicts the occurrence of something else (for example, the ringing of a door bell predicts that someone is waiting at your front door for your attention).

Pavlov (1927) was the very first to systematically address this question. He suggested that if an animal could learn arbitrary associations, such as the ringing of a bell and the presentation of food, then the animal was merely learning contiguous relationships and not contingent ones. Pavlov further suggested that the association between a stimulus and a response was describable as a physical one, i.e. they are linked via 'nervous connections'. This type of learning has become known as classical or Pavlovian conditioning. Other forms of conditioning are known as operant or instrumental conditioning.

Classical conditioning is an important way of learning responses in new situations. Under normal circumstances, when a stimulus of biological significance such as food is presented to an animal, some behaviour or reflex will be immediately evoked, such as salivation. Pavlov (1927) first observed that by repeatedly ringing a bell immediately prior to feeding time, dogs would begin to salivate before the food became visible. He speculated that ‘nervous connections' between the sensory registration of the ringing bell and the salivatory response were formed during this repeated experience, enabling an ‘anticipatory' response to be learned. Such a finding provoked further investigation, not only by Pavlov, but also by many subsequent researchers. Such work has uncovered a wealth of related phenomena that were thought to be generalisable across species, including humans.

In classical conditioning, it is said that when an unconditioned stimulus (UCS) such as food is presented to an organism, an unconditioned response (UCR) such as salivation is evoked. The repeated pairings of a conditioned stimulus (CS), say the sound of a bell, and a UCS result in the formation of a conditioned response (CR) with the occurrence of the CS alone, that is to say, salivation at the sound of the bell.

Pavlov pioneered research into:

•  the acquisition of CS–UCS associations;

•  the unlearning of associations, known as extinction;

•  the ability of the CS to form further associations with other previously neutral stimuli, known as second-order conditioning;

•  the effects of altering the time interval between onset of the CS and onset of the UCS (the interstimulus interval, or ISI);

•  the effects of conflicting associations (Pavlov was one of the first to develop a theory of neurosis based on conditioning principles).

Acquisition and extinction

Pavlov (1927) found that the acquisition rates of CS–UCS associations could be replicated using different animals, which suggests that such rates may follow a general law. The rate of acquisition follows an S-shaped, or ‘sigmoid' curve where the vertical axis represents the magnitude of the response (a measurement of the frequency or intensity of the response, say the number of drops of saliva) and where the horizontal axis represents the stage of conditioning (usually the trial number). This curve indicates that learning is initially slow but is shortly followed by a period of rapid learning. This high rate is then tempered by a gradual rise to a peak.

When the learned CS is presented alone, i.e. it is presented in the absence of the UCS, the magnitude of the CR decreases until eventually it disappears. However, it is sometimes observed that during the latter phases of the extinction procedure the CR suddenly, but briefly, reappears. The cause of this reappearance, known as ‘spontaneous recovery', is not well understood.

Generalisation and discrimination

Suppose some noise such as a tone was used as a CS in a conditioning experiment. Suppose further that this tone was pitched at ‘middle C'. If the pitch was altered slightly while testing for a response, would the CR be observed? Results of such tests reveal that the closer the similarity of a novel stimulus to the CS, the more equivalent is the magnitude of the CR. For example, a CR trained to a CS consisting of a tone of 1,000 Hz generalises within the range 800 to 1,200 Hz.

Generalisation is important for all animals, including humans, since the variance of a single stimulus can be so great that without this ability every variation, however small, of a single stimulus would need to be learned. Clearly, this is a vital capacity as it is rare that the exact same stimulus will be encountered twice.

Likewise, it is equally important to know where this generalisation boundary should end, that is discriminating between biologically significant and non-significant variations of a stimulus.

Temporal contact, or contiguity in time, refers to the fact that close contact in time between the CS and the UCS seems to be essential for learning to occur.

The temporal relationship between the CS and UCS is studied by varying the time between CS onset and UCS onset and observing the resulting conditioning rates. It is possible to identify four variants of temporal contact that produce reliable differences in the rate of learning:

•  Trace conditioning. This is where the CS precedes the UCS but is terminated prior to or at UCS onset; acquisition rates are high (see Figure 4.1).

•  Delay conditioning. This method produces the highest acquisition rates; this is where the CS precedes the UCS but stays on with the UCS.

•  Simultaneous conditioning. Very weak conditioning occurs during simultaneous conditioning, where CS onset and UCS onset coincide.

•  Backward conditioning. No conditioning is observed during backward conditioning, where the UCS precedes the CS.

If the duration of the CS and the duration of the UCS are held constant, then trace conditioning displays the longest ISI, delay conditioning will have a shorter ISI, while zero ISI implies simultaneous conditioning and negative ISI implies backward conditioning.

The findings of ISI studies were initially difficult to explain since it was thought that it was the simultaneous pairings of the CS and the UCS that resulted in learning. Hence optimal ISI was expected to be 0 ms or thereabouts. Explanations as to why these various temporal contact procedures should be ordered so, in terms of rate of CR acquisition, concern the predictive nature of what is learned in classical conditioning. Animals learn that the CS predicts the UCS, and hence this learning is more rapid when in classical conditioning experiments, the CS precedes the UCS.

Contingency and ‘blocking'

It has been suggested that classical conditioning is the learning of a predictive relationship between the CS and the UCS (Kamin 1969, Rescorla and Wagner, 1972). In other words, the strength of the CS–UCS association increases with respect to the probability of their concurrence, and is contingent upon the CS being a reliable predictor of the UCS (Rescorla, 1968). Close temporal presentation of the CS and UCS, it is assumed, results in the increase in the associative strength between their neural representations (Hebb, 1949; Rescorla and Wagner, 1972).

The idea of contingency, i.e. that the CS is a predictor of the UCS, is highlighted by the ‘blocking' effect first reported by Kamin (1969). A control group of animals is conditioned to evoke a CR to two simultaneous CSs (CS1 and CS2, which are followed by a UCS). On presentation of CS2 alone, the CR still occurs. However, the blocking group is initially trained to produce the CR to CS1 only, before being subjected to the control procedure (CS1 + CS2, followed by the UCS). Subsequently, when tested, the blocking group shows significantly fewer CRs to the presentation of CS2 when it is presented alone. Thus, after both procedures, the blocking group shows a smaller CR to CS2. A reasonable explanation is that for the blocking group the CS1 was sufficient to predict the onset of the UCS. Subsequently, when exposed to CS2, it is in effect ignored. This study shows that learning is based on contingency, that is whether or not a stimulus predicts another event. In the case of the blocking group, CS1 may predict the UCS more reliably than will CS2.

Secondary or higher order conditioning

Once a CS–UCS association has been learned then it is possible to form an association with the CS and a second CS even in the absence of the UCS. The first CS can act as a UCS for a new CS. Again, when the second CS reliably produces the CR, it too can be used to condition a third CS. Secondary conditioning is learning in the absence of a UCS, and where previously neutral stimuli may acquire both the power to form further associations with other neutral stimuli.

Hull (1932) was the first to suggest that higher order conditioning may be responsible for the learning of complex skills or ‘habits'.

Many students when writing about learning theory, merely provide answers to the question: what do you know about classical conditioning? Hence they write down everything they can remember about the topic. I have seen many such essays receive a fail grade. You must address the question. Typically, something very specific will be asked, such as the role of contingency or contiguity in learning.

Inhibitory conditioning

So far an established CS has been considered as being excitatory, that is the CS representation excites the UCS or CR representations. In such cases it may be said that the CS signals the imminent arrival of the UCS. However, a CS may also signal the omission of the UCS. A CS that does this is known as a conditioned inhibitor. For example, pigeons learned to move away from a light when food was only delivered while the light was off (Hearst and Franklin, 1977). Another demonstration of conditioned inhibition is described by Zimmer-Hart and Rescorla (1974) in which the CR was suppressed during the combined presentation of a CS1+CS2, but at its maximum during only CS1 presentations. This suggests that CS2 inhibited the CR. It is usual to suppose that conditioned inhibitors form a negative association with the CR (denoted CS”).

Inhibitory conditioning is a useful concept because not only can stimuli predict the occurrence of something, they can also predict the omission of something. Suppose you have a friend, person A, who is nearly always very pleasant to you. She laughs at all of your jokes and generally makes you feel good. However, when she is accompanied with another friend, person B, she is very cool towards you and shows you none of these positive attitudes. After some time, you will come to predict that person A plus person B equals the omission of pleasant exchanges. Yet you will predict that person A alone equals the presence of positive exchanges. Thus person B is an inhibiting stimulus and you would learn not to seek out their presence.

Pre-exposure effects

When a stimulus is repeatedly presented to an animal and this stimulus has no obvious consequences, i.e. it is not regularly followed by a UCS or an acquired CS, then later attempts to condition this stimulus are disrupted (Baker and Mackintosh, 1977). It is common to say that repeated exposure to a stimulus that does not signal any specific event results in the reduction of a response to that stimulus. This process is known as latent inhibition, since it inhibits future learning. Sensory preconditioning, on the other hand, is the exact opposite of latent inhibition. When two neutral stimuli repeatedly occur together, then the conditioning of one of the pair to a CS enhances the conditioning of another. For example, if neutral stimuli CS1 and CS2 occur together, then training CS1 with a UCS results in more rapid conditioning for CS2 when it is later paired with the UCS. Pre exposure effects seem to indicate that no reinforcement (association with a UCS or established CS) is required for conditioning.

Operant conditioning

Whereas in classical conditioning responses are typically reflexive (salivation, hand withdrawal from heat and the eye blink reflex, etc.), animals display other behaviours that are not necessarily elicited by specific stimuli. These spontaneous responses, or what Skinner referred to as operants, are species-typical behaviours (such as pecking, pawing, running, moving the head, and so on). Stated simply, operant conditioning is learning that an operant or action may have specific consequences. This is in contrast to learning when one event signals another as in classical conditioning. Viewed in this way, operant conditioning is, then, a more active process than classical conditioning.

Thorndike (1911) pioneered work in this paradigm, postulating the Law of Effect. This states that actions which are followed by some form of ‘satisfying' stimulus are more likely to occur in the future, and that the greater the level of satisfaction the greater the strength of the bond between the response and its believed consequences. Thorndike defined satisfaction as a stimulus that does not invoke avoidance but rather one that invokes behaviour designed to attain it. The word ‘reinforcer' has come to replace ‘satisfying stimulus', and is used to denote objects or events that modify behaviour in some way.

A criticism of the concept of reinforcement is that a reinforcer is only defined by its effects and we cannot predict them accurately in advance. For example, suppose that psychologists believe that a slap in the face is a negative reinforcer. But if a barmaid slaps me in the face and then I frequent the pub more often then the slap has to be regarded as a positive reinforcer. In other words, only the effects of a stimulus can tell us whether it is a positive, negative or no reinforcer.

Primary reinforcers may include desirable stimuli or outcomes (positive reinforcers) as well as aversive stimuli or outcomes (negative reinforcers). It is assumed that responses are continuously emitted, perhaps being randomly selected ‘until the right one shows up' (Skinner, 1969, p. 134). That is, a range of behaviours is made until a desirable state in the immediate environment occurs. Such responses are said to be contingencies for reinforcement. For example, in the ‘conditioning chamber' the experimenter rewards particular responses; in real situations some behaviours are more likely to lead to the discovery of desirable stimuli, such as food, than others are. In both situations the probability that these responses will recur is increased.

Most of the early work on operant learning was conducted on rats learning to find food in mazes. Indeed, it was Tolman's belief (a major contributor to this field) that everything important in psychology could be reduced to the processes involved in a rat at a decision point in a maze. While this may be rather overoptimistic, some very important issues have been raised by animal maze experiments. However, as well as a number of reliable concepts and sophisticated theories, a few controversies and theoretical problems have emerged from such work.

Place learning

Animals such as rats are especially good at learning the locations of reliable sources of food. This ability is fully exploited by psychologists interested in the process of place learning, especially the kinds of mental associations that might be formed. Are the responses of a rat finding a goal location in a maze a sequence of chained responses, so that one response automatically triggers the next one, or does it learn its environment as some mental or cognitive map?

Maze learning involves learning to coordinate serial behaviour towards the goal state. Many types of other skilled behaviours involve learning to do precisely this (see Chapter 7 on the state space approach to problem-solving). One of the first findings in the paradigm of operant conditioning was the discovery that the time lapse between the moment of reinforcement and the action responsible for it is crucial, but that such time intervals could be much longer than in classical conditioning. When this time lapse is minimal conditioning is most efficient. However, conditioning with delays of up to twenty minutes are still possible.

In an attempt to explain how conditioning with such large time delays was possible, especially in maze learning, Hull (1943) developed the goal gradient hypothesis. He supposed that finding the goal box of a maze was experienced as reinforcing for the rat. This reinforced the move into the goal box, thereby making this response at this point in the maze more likely to occur in the future. It also reinforced the association between the goal location and the food found there. This latter effect is crucial to the theory for it suggests that on future trials the location visited immediately prior to the goal box would also come to be associated with food (through secondary conditioning). This ‘chaining' of the responses made at each location would eventually be extended backwards to the start location.

Evaluation

Despite this apparent success, response chaining has a number of problems, and from these the competing cognitive map hypothesis emerged. A major premise of Hull 's theory is that action sequences are formed as a set of chained responses. However, a number of studies have questioned this assumption. For example, Tolman (1936) found that when a learnt maze (Figure 4.2(a)) was redesigned with straight alleys that went off in different directions (Figure 4.2(b)), rats tended to choose the alley, which was closest in direction to the goal location (route B in Figure 4(b)), rather than closest to the first choice point of the original design (route A in Figure 4.2(b)). If responses were chained, then the rat should have chosen that path from the start location nearest to the original path (route A). To account for this behaviour, Tolman suggested that animals utilise environmental cues or landmarks to develop a cognitive map of the environment – a mental representation of the relationship between different landmarks.

Figure 4.2 The maze used by Tolman (1936). When the learned route was blocked, rats chose the route that led directly towards the light (B) rather than the route closest to the learned route (A), indicating that the rats learned a spatial map of the environment rather than a chain of responses

The cognitive map hypothesis

More recent studies have demonstrated that rats do use external cues (external to the maze) such as filing cabinets or windows in learning about mazes. For example, rats were placed in a circular pool of milky water where a small, dry platform, not visible at the surface level, was placed in a fixed location (Morris et al., 1982). Since rats dislike being placed in water, they swam around until they found the platform, whereupon they rested. After being repeatedly placed in the pool, they learnt to swim directly to the platform. When placed at new points in the pool the rats still swam almost directly towards the platform. Experiments in animal place learning such as this have shown that if all external cues are eliminated, then the probability of finding the goal is reduced to chance levels. Taken together, Morris et al. (1982) argued that this is evidence that rats use a cognitive map.

Evaluation

The cognitive map is not compatible with behaviourist assumptions. It is not about response chaining but having an internal model of the world, a term that the Behaviourists were keen to avoid. It is now generally accepted that the cognitive map theory is a better explanation of place learning than response chaining.

Schedules of reinforcement

A different approach to operant learning was adopted by Skinner. Rather than study animals learning their whereabouts in a maze, Skinner focused on the frequency and timing of reinforcement in order to examine the strongest and weakest forms of learning (acquisition) and unlearning (extinction). Skinner's main findings are that the power of reinforcement depends upon whether it is given over a fixed or varied interval, or whether it is contingent upon a particular frequency of responding. There are two main types:

•  Continuous reinforcement. The simplest (but most unrealistic) form of reinforcement is to provide it every time the desired response is made.

•  Partial reinforcement. Reinforcement is not continuous but depends on the number of responses or the amount of elapsed time.

These are further subdivided into the following:

•  Fixed interval (FI)

Reinforcement can only occur after a fixed interval of time and when after the interval the desired response is made. For example, an animal might only be reinforced on the first appropriate response after one minute. An example of fixed interval is a wage slip at the end of the week or month, provided at least once when response is made (i.e. turning up to work and doing the job). Skinner found that the animal initially responds at a very low rate but this rate soon picks up. Behaviour shows a cyclic pattern where responding is highest near the time of reinforcement (known as the FI scallop). Because of this scallop, it is not a good schedule for encouraging continuous responding. An effect of this you may be aware of is the scheduling of exams, e.g. at the end of term or semester. This predicts that the peak of responses will occur near the moment of ‘reinforcement', that is students will cram in their revision a week or so before the exam.

•  Variable interval (VI)

Reinforcement occurs after a specified interval of time and when the desired response is made. The interval can change from trial to trial, thus it can be 5 seconds on one trial and 2 minutes on the next. An example of this is reading your e-mail. You click on ‘Read Mail' and your response is rewarded over variable intervals of time (sometimes you won't get an e-mail for a couple of hours, but at other times they might be seconds apart). Responding on this schedule appears consistent throughout the trials (i.e. a straight line on the graph). This is a better way of scheduling exams (i.e. when the students aren't told which week they will be in!), since the students will work continuously.

•  Fixed ratio (FR)

Reinforcement depends upon the number of responses made regardless of time. Animals show consistent responding on this schedule. However, if the ratio requirement is too high animals show ratio strain, which is characterised by long pauses in responding. The student equivalent of this might be if a course has a large number of assignments then the student might lose interest in studying, and show long periods of avoiding it. The reason is that a lot of work has to be done before the reward is given.  

•  Variable ratio (VR)

Reinforcement depends on a number of responses, and this number is varied on a trial-by-trial basis. Like VI schedules, VR results in responding that is consistent over time. An example of a VR schedule is the fruit machine. It pays out only after a certain number of coins have been inserted and this number is variable.

Schedules of reinforcement and extinction

The process of extinction is the attempt to eliminate the response by no longer giving reinforcement. For a continuous schedule, the extinction process works very quickly. This is not surprising: if the reinforcement is highly predictable then responses will cease when it is no longer reinforced. Behaviour acquired through partial reinforcement is more difficult to extinguish (e.g. Lewis and Duncan, 1956). Since reinforcement was less predictable anyway, its omission is less surprising so responding will continue longer than when it is continuously reinforced. This can explain, for example, why fruit machines can be so addictive: if the machine never paid out then you would soon stop playing. However, since it pays out on a partial reinforcement schedule (VR in fact) then it is more difficult to stop. Avoidance behaviour is difficult to extinguish and this is because the behaviour is always reinforced (the potential threat is not experienced).

Methods of reinforcement provide ways of establishing a range of desired responses. However, in some cases the desired response might be too specific or the response made too infrequently to reinforce directly. The method known as shaping is used to reinforce related behaviours rather than the specific behaviour itself. For example, suppose a rat is to be trained to press a lever. Since lever pressing might not be part of the animal's usual behaviour, the experimenter reinforces the rat for moving in the direction of the lever. Once this behaviour is acquired, then the experimenter only reinforces the rat when it is within a few inches of the lever. Subsequently, when the rat has learned that being close to the lever is a good way of obtaining food, the experimenter only reinforces the rat when it touches the lever. In shaping, then, the reinforcement becomes more specific as behaviour approaches the desired response.

Research on these schedules enabled Skinner to achieve some remarkable behaviour in animals. For example, through selective reinforcement pigeons were able to play a sort of table tennis game, and the story goes that the US military explored the use of pigeon guided missiles. The idea is that the pigeons are trained to peck certain keys in the presence of a particular target. This way, the pigeon, perched inside a missile and looking out through a small window, would be able to guide itself to the desired target.

Evaluation

Skinner generalised these findings to human behaviour. Indeed, it has been argued that he over generalised his findings in his claim that all behaviours are acquired through contingent reinforcement, from learning language to learning bad behaviour. Despite these criticisms, learning theory has been applied in a wide number of settings. In the next section we examine the legacy of learning theory by looking at how it has been applied to human behaviour.

 ection 3

Applications of learning theory

In this section we examine applications of learning theory to human learning, and in doing so we will evaluate the usefulness of the approach in understanding human behaviour and human problems.

Can humans be conditioned?

Before one can apply learning theory to humans, we first need to establish that humans can be classically conditioned and that human learning can be guided by contingent reinforcement.

Early demonstrations of human conditioning showed that:

•  the eye-blink response can be conditioned to a tone (Spence, 1969);

•  skin conductance can be conditioned to a tone (Öhman et al., 1975);

•  human judgement can be classically conditioned (Levey and Martin, 1975);

•  blocking effects are also observable in humans (Trabasso and Bower, 1968);

•  conditioning principles may be at work in advertising (Gorn, 1982; Smith and Engel,1968);

•  reinforcement principles may be used to modify behaviour in schools and prisons (Hall et al., 1968; Phillips, 1968).

Some have claimed that classical conditioning might also occur on the psycho-sociological level. For example, Turkkan (1989) claims that personal attributes may be transferred by association. For example, person A (the CS) may take on attributes associated with person B (UCS) by being seen frequently together (paired). Recent studies show that such conditioned learning can be developed in the lab (Baeyens et al. 1990; Fulcher et al., 2001).

An interesting study by Smith and Engel (1968) also demonstrates that conditioning can also occur without conscious awareness that it is taking place. Photographs of cars with or without the presence of a very attractive woman standing next to the car were shown to a group of male subjects. Despite the denial in almost all participants that the attractive women had influenced their ratings of the desirability of each car, their evaluations were much higher for those cars accompanied by attractive women.

Another study of this effect is that by Levey and Martin (1975) on postcard evaluation. Fifty scenic postcards were presented to human subjects who were asked to rate them according to how much they liked or disliked each picture. For each subject the two most liked and two least liked were selected for the second part of the experiment. Each of these cards was paired with a neutral card (valued as neither liked nor disliked) and presented on a viewing screen. A neutral card represented the CS while a liked/disliked card represented the UCS. Each was paired according to a standard trace conditioning paradigm. In the third part of the study, subjects were asked to rate the neutral cards again. It was found that their ratings of the neutral cards had been affected by whether they had been paired with liked or disliked postcards. Neutral postcards paired with liked postcards themselves became more liked than neutral postcards paired with disliked postcards.

Operant conditioning theory of depression: learned helplessness

In an operant conditioning experiment first reported by Seligman and Maier (1967), animals were discovered to learn to cease producing responses that would otherwise have avoided a shock. The experiment involves two phases and three groups of animals. In the first phase, all animals were placed in hammocks. Group Escape were given electric shocks that could be avoided by pressing a panel with the nose, Group Yoked were given the same number of shocks as Group Escape but were unable to avoid the shocks, and Group Naïve received no shocks. In the second phase, each animal was placed in a chamber in which an occasional shock was delivered. If the animal moved to the other end of the chamber, the shocks ceased. Within a few trials Group Escape and Group Naive learned to avoid the shocks. However, Group Yoked was unable to learn to avoid the shocks. Instead they had learned that they were ‘helpless'. This study and many variants of it have formed one basis of theories of depression in humans, by supposing that depression is a reaction to a series of events in which the adverse consequences were believed to be outside the person's control. In order to explain why some people get depressed and others do not under the same circumstances, Seligman introduced the notion of attribution. For example failing an exam might be attributed to external factors (‘The lectures did not prepare us for this exam') or internal ones (‘I'm hopeless at exams'). According to the theory, individuals who attribute failure to internal sources are more likely to become depressed.

Classical conditioning theory of phobias

Intense fear of a particular object or situation may have been acquired through classical conditioning. A CS is paired with an extremely aversive and fearful event or stimulus (the UCS). Subsequently, the CS becomes feared and is avoided. An example might be a child who cuts itself on barbed wire (UCS) while stroking a rabbit (CS). The child then learns to fear rabbits.

One problem with the conditioning theory of phobias is that the things that people fear tend to have some basis in rationality. For example, fears of snakes and spiders might merely represent what our ancestors feared, and indeed some varieties of snakes and spiders are harmful. Phobias of inane objects, such as socks, sticky tape, and grass and so on, are rarely reported. Nevertheless, there is evidence that intense fears can be acquired through classical conditioning (Davey, 1992).

Behaviour therapy

If some psychological disorders are acquired through conditioning, then it should be possible to eliminate them through conditioning-based procedures.

Several of these have been designed, such as the following:

•  Systematic desensitisation

The principle of counter conditioning can be used to associate the feared object with something pleasant. For example, Jones (1924) first used this method to eliminate fear of rabbits in a young boy, pairing the rabbit with eating. The rabbit was gradually introduced over a period of days each time the boy was eating. Wolpe (1958) coined the term systematic desensitisation and used it by getting the individual to imagine various feared objects and situations during periods of deep relaxation, thus associating the feared stimulus with a relaxed state.

•  Aversion therapy

Rather than cure fears, aversion therapy is aimed at curing addictions or unwanted habits such as smoking. The technique of counter conditioning is used to associate behaviours involved in the habit with something extremely aversive, such as vomiting. Raymond (1964) used this procedure to help a 14-year-old boy quit smoking. The boy was given injections of apomorphine, which induces nausea, while he was smoking. Subsequently, the boy felt quite ill the next few times he tried to light up. This lasted for several years, apparrently.

•  Classical conditioning for enuresis

Mowrer (1938) developed a treatment for enuresis (bed-wetting) that is based on classical conditioning principles. Moisture detecting equipment is attached to the bed and when the bed becomes wet the equipment sets off a loud buzzer that wakes the child. The claim is that the sensation of a full bladder (CS) becomes associated with the buzzer (UCS) and hence the child wakes (CR) to the feelings of a full bladder. Studies have shown this to be a very successful technique (Doleys, 1977). Unlike many other methods of therapy, behaviour therapy treats a disorder as merely unwanted behaviour. It is not assumed that the disorder is the result of some inner conflict (e.g., as in psychoanalysis) or the consequence of a dysfunctional family and so on. A more dominant approach in clinical psychology recently is cognitive-behaviour therapy, in which counterconditioning principles are used in tandem with other techniques based on cognitive principles.

The token economy

There are numerous reinforcers that are used socially. Perhaps the most common one is praise, and giving praise for certain behaviour usually results in an increase in that behaviour. For children, attention can be the strongest reinforcer and it seems that they will do anything (including behaving badly) in order to attain it. However, there are settings where praise and attention can act as poor reinforcers, such as with prisoners, disruptive children or individuals with severe learning difficulties. An alternative is the token economy, in which points or tokens are given for desirable behaviour and these can then be exchanged for sweets or other goodies. At a residential centre called Achievement Place juvenile delinquents were given points for appropriate behaviour and these could be exchanged for snacks, money or special privileges. There were significant gains in good behaviour, including a major increase in the amount of homework completed (Phillips, 1968). Token economies have also been used successfully in the workplace to reduce the frequency of accidents (Fox et al, 1987). The method has also been successful in prisons by increasing the amount of time prisoners spent improving themselves through study, and in one case significantly raising IQ (Kandel et al., 1976).

Problems with the token economy include the following:

•  Extinction

When the token economy is terminated, behaviour may return to its ‘pretoken' form (Wolf et al., 1987).

•  Tokens as bribes

One objection to the token economy is that the reinforcer is nothing other than a bribe. However, as Lieberman (2000) argues, society is very welcoming of negative reinforcers, such as prison itself, being ‘grounded', being slapped and so on, and all of these are unpleasant (as well as ineffective): ‘†reinforcement [such as the token economy] can bemore effective than traditional forms of discipline in at least some circumstances, and avoid harmful side effects that sometimes come with punishment' (p. 262).

•  Extrinsic motivation

The desired behaviour is motivated towards obtaining tokens or points rather than through an internal desire to behave in more socially acceptable ways. However, the benefits of changing behaviour can often outweigh the negatives.

Learning theory and child rearing

Many parents use punishment as a way of controlling their child's behaviour. But the evidence that punishment works is not impressive. The negative effects of punishment include the following:

•  Fear

Since punishment involves the use of aversive stimuli, and aversive stimuli can induce fear, then the risk is that punishment over a long term can increase fear and anxiety. Increased anxiety can also, in turn, hamper the ability to pay attention (Cheyne et al., 1969).

•  Avoidance

If a child is punished for not achieving a certain standard on a specific task then rather than try harder the child might avoid the task altogether (Martin, 1977).

•  Increases in aggression

The pain associated with punishment can under some circumstances promote aggression. Furthermore, a child brought up in an environment where physical punishment is the normmay learn to see aggression as a normal way of behaving.

Alternatives to punishment

According to the principles of conditioning there are several effective ways of encouraging desired behaviour and minimising unwanted behaviour without recourse to punishment:

•  Withdrawal of attention

Children often seek the attention of others, especially parents and teachers. When a child is misbehaving, carers are naturally tempted to attend to the child. However, the attention may be rewarding and hence can reinforce the behaviour. An alternative is to ignore the child and hence not provide a reward for the behaviour. However, this is only likely to work for minor misbehaviours, and advocates such as Wierson and Forehand (1994) suggest including withdrawal of attention in conjunction with reinforcement for good behaviour.

•  Reprimands

Carers often respond to a misbehaving child by ‘telling them off' or reprimanding them. Often reprimands are used inconsistently, are shouted out and result in argument. O'Leary (1995) has studied the use of reprimands and claims that they work best when they are:

•  consistent – if a child is told off once for a particular behaviour then the child is always told off for doing it;

•  immediate – the time between the behaviour and the reprimand should be minimised;

•  brief – often carers get involved in arguing about a particular behaviour. The advice is to not engage in such arguments and to keep the reprimand brief;

•  firm in tone – the use of anger is claimed to be unnecessary and undesirable. A firm tone is better than a yell.

•  Time-out

The misbehaving child is removed to place that is less reinforcing, such as sitting in the corner of a room watching other children play but not being allowed to join in (White and Bailey, 1990). Time-outs used in the home such as ‘go to your room' may be ineffective since the child's room is a reinforcing environment. Research shows that time-outs work best when they are brief (one to two minutes) and when the child is gently constrained.

Typical Examination Questions

1. Outline how contingency and contiguity are important for learning in classical conditioning.

2. How do learning theorists explain how sophisticated behaviour (such as spatial knowledge, addiction, skills and so on) is acquired?

3. Outline and evaluate one application of learning theory.

Further reading

I can only suggest one text for learning theory, and it is the superbly written:

Lieberman, D. A. (2000) Learning, Behaviour and Cognition, 3rd edn. Belmont , CA : Wadsworth Thomson Learning.

Next >>

This book was first published in 2003 by Crucial, a division of Learning Matters Ltd [ISBN 1 903337 13 5] © 2003 Eamon Fulcher; © 2009 GEFT Consultance Services (geft.co.uk).

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior permission in writing from Geft Consultancy Services, who may be contacted via www.geft.co.uk.