What is the initial learning of the connection between the unconditioned stimulus and the conditioned stimulus when these two stimuli are paired?

3.1 Learning

Classical conditioning involves pairing an unconditioned stimulus (UCS), e.g., a charismatic, enthusiastic teacher who emphasizes the excitement and importance of a topic, with a neutral stimulus (a not-particularly exciting curriculum) which becomes the conditioned stimulus (CS) and evokes enthusiasm and dedication in the student. Pairing morphine (UCS) given to patients in great pain, causes the nurse and syringe to become conditioned stimuli (CS) for pain reduction. Placebo effects persist when only the CS is present and, following principles of partial reinforcement, continue undiminished if the UCS is reintroduced periodically, but eventually extinguish if the UCS is never reintroduced. For example, if a painkiller (UCS) having severe side-effects is interspersed with a placebo in an identical capsule (CS) and intermittent reintroduction of the UCS, the effect may be a continuously effective analgesic with diminished side-effects. There is a powerful relationship between magnitude of UCS and response to the CS. For example, a massive dose of morphine (UCS) will result in a massive placebo (CS) effect. These principles of conditioned analgesia have been found with animals as well as humans.

Ethics requires disclosure of use of conditioning; e.g., following explanation of learning theory, the doctor might state that the UCS would be intermittent and that the client's own relevant powers will serve to reduce pain. Conditioning produces expectancy, and memory of UCS effectiveness, as well as measurable neurological change.

2.2 Occasion Setters are Impervious to Extinction

Since simultaneous FP discriminations are solved by the model almost completely by direct X–US associations (see Fig. 1), the effect of X extinction is dramatic, abolishing CR responding on both X and XA test trials. By contrast, compound responding for serial FP is controlled by A–US associations which are attenuated by the inhibitory H–US associations (see Fig. 2). Since X–H and H–US associations are only minimally affected by X extinction, very little decrement is observed in XA compound responding after extinction.

3 Conclusion

With classical conditioning of the Aplysia withdrawal reflex, the paired CS and US form an association by converging on a second messenger cascade within a single cell. This convergence results in the enhancement of a specific synapse. With operant conditioning of Aplysia feeding behavior, the association is made through contingent reinforcement. Contingent reinforcement of the response results in the alteration of a cell that mediates the expression of that response. Conditioning occurs through a modulation of the membrane properties of this single cell. Thus, modifications made to individual neurons (via intrinsic membrane properties and synapses) can account for both types of associative learning phenomena.

2.1 Habituation

Habituation is perhaps the simplest form of learning. In the absence of an unconditioned stimulus, the response to any conditioned stimulus degrades with each repetition of that conditioned stimulus. In biological neural systems, it has been observed that neurons respond more strongly to stimuli which occur infrequently. If a stimulus occurs often and is not classically conditioned, the neuron loses its ability to respond to that stimulus. Conversely if the stimulus is not observed for a long period of time, the neurons ability to respond may return. Experimenting with Aplysia has clarified the neural basis for habituation[BK85]. Bailey and Kandel showed that habituation was localized at a single synapse. Repeated firing of the presynaptic neuron depressed the strength of the synaptic connection. The postsynaptic neuron, therefore, responded less strongly to any stimulus which activated the presynaptic neuron. This behavior showed both short and long term effects. In the short term, synaptic strength could be caused to decrease quickly and then rebound. Conversely, if the synaptic strength was reduced for a long period of time it required a long period of time to reestablish itself. Short term activation of the presynaptic neuron was found to reduce the influx of Ca2 + which is necessary for neurotransmitter release. Long term habituation led to long periods of Ca2 + deprivation which in turn made the electrical connection between neurons immeasurable and caused changes in the physical structure of the synapse.

The Byrne and Gingrich model is based on the flow of neurotransmitter among the external environment and two pools internal to the neuron [BG89]. One of the pools, the releasable pool, contains all the neurotransmitter ready to be released when the neuron is activated. The other pool, the storage pool, contains a store of neurotransmitter for long term use. Short term habituation is explained as the depletion of the releasable pool due to frequent activation of the neuron. The increased level of Ca2 + which results from the occurrence of the conditioned stimulus increases both the flow from the storage pool to the releasable pool and the release of neurotransmitter. It is the increase in neurotransmitter release which leads to activation of the neuron. In this manner, both pools are depleted by the frequent occurrence of conditioned stimuli. The neurotransmitter in the releasable pool can be replenished by diffusion from the storage pool or by other neurotransmitter flows which are regulated through sensitization and classical conditioning. Long term habituation can be explained by depletion of the storage pool itself. The Byrne and Gingrich model assumes a single flow into the storage pool from the external environment. This flow is proportional to the difference between the current neurotransmitter concentration in the pool and the steady state value.

Another model for both long term and short term habituation was presented by Wang and Arbib and is reproduced as follows with straight forward modifications[WA92]. This model was produced independently from the Byrne model, and is based on Wang and Arbib’s own experimental observations. The modifications introduced here merely change the equations from continuous to discrete time, which is necessary in order to use them later in an artificial neural network. Here I(t) is the current input vector from the neurons whose outputs are habituated, and W(t) is the vector of synaptic strengths. The dependence of W(t) on sensitization or classical conditioning effects is ignored. Since this habituation only version of W(t) is dependent only on the activation of the presynaptic neuron, only the single subscript, i, is needed. Synapses attached to the same presynaptic neuron habituate in the same manner. Henceforth W(t) will be referred to as the habituation value to avoid confusion with either a more complete model of synaptic strength or artificial neural network parameters.

(1)Wit+1=Wit+τiαzitW i0−Wit−WitIit

(2)zit+1=zit+γ zitzit−1Iit

In this model, τi is a constant used to vary the habituation rate and α is a constant used to vary the ratio between the rate of habituation and the rate of recovery from habituation. The function zi(t) monotonically decreases with each activation of the presynaptic neuron. This function is used to model long term habituation. Due to the effect of zi(t) after a large number of activations of the presynaptic neuron the synapse recovers from habituation more slowly. Some assumptions about the range of values for the constants are made in order to assure that Wi(t) and zi(t) remain within the range [0,1]. Specifically, τi, γ, the product of τi,·and α and Ii(t) must all be in the same range [0,1]. For simplicity, Wį(0) will always be assumed to be unity unless otherwise stated.

It is apparent in Wang and Arbib’s model that in the long term, if the presynaptic neuron is not completely inactive the synaptic strength will eventually decay to zero, because zi(t) is monotonically decreasing. This was valid for Wang and Arbib’s research because they were examining the response of animals to artificial stimuli which are of no importance to the animals in question. Sensitization and classical conditioning are ignored in this model. If these other two learning mechanisms were included in the model, then the synaptic strength would only decay to zero in the absence of any unconditioned stimuli.

1 Basic Terms

The pairing of an initially neutral stimulus (the conditioned stimulus—CS) with a biologically relevant stimulus (the unconditioned stimulus—US) comes to elicit a response (conditioned response—CR) that is usually but not always similar to the response previously associated with the unconditioned stimulus (the unconditioned response—UR). In fear conditioning, the US is an aversive fear-eliciting stimulus such as painful electric shock or loud noise, the CS is a neutral tone or light stimulus. The unconditioned and the conditioned response consist of changes on the subjective, the behavioral and the physiological level and include (in humans) enhanced subjective fear and responses such as freezing, changes in heart rate and skin conductance, the release of stress hormones, reduced pain sensitivity and startle reflex potentiation.

The development of the CR is based on the formation of an association between a neutral stimulus and a stimulus with innate biological significance (Rescorla 1988). Most studies involving fear conditioning have used cue rather than context conditioning, i.e., discrete CSs were presented rather than using the environment of the animal (e.g., the cage) as CS. In addition, delay conditioning where the CS terminates with the US rather than trace conditioning where the CS and US are separated in time were used in most studies. Fear can be viewed as a specific reaction to threatening stimuli. It can turn into an anxiety disorder when the fear becomes disproportionate to the stimulus that elicits it or when fear is experienced in inappropriate situations.

4 Clinical Interventions Involving Operant and Respondent Conditioning

Respondent (also known as classical, or Pavlovian) conditioning (see Classical Conditioning and Clinical Psychology) involves the pairing of unconditioned and conditioned stimuli, which ultimately leads to a conditioned stimulus that elicits a conditioned response. One of the more useful clinical heuristics has been research on how respondent and operant conditioning can combine to explain important clinical problems.

The best-known problem that has been addressed by considering both operant and respondent conditioning is the theory of the acquisition and maintenance of phobic behaviors. It has been suggested that phobic behaviors are acquired by classical conditioning but maintained by operant conditioning. Consider the simple example of someone bitten by a dog. In respondent conditioning terms, the dog bite is an unconditioned stimulus that produces the unconditioned response of pain and fear. Following such an incident, the next time the person approaches a dog, their fear and anxiety rises as the stimulus (the dog) gets closer. So far, the acquisition of the fearful response can be understood using a classical conditioning paradigm. If the person were to approach a variety of dogs, the fearful response would extinguish naturally, because extinction in classical conditioning is accomplished by presenting the conditioned stimulus (a dog) in the absence of the unconditioned stimulus (the dog bite). If this were the case, phobic responses would extinguish naturally over time. However, in many instances when one sees the dog and anxiety increases, a person simply turns around and walks away, thus avoiding the feared object. When that happens, the avoidance behavior is negatively reinforced (increased) by the removal of the anxiety. This increases the probability of avoiding the dog the next time such a stimulus is encountered. The avoidance of the phobic object prevents the natural extinction of phobic anxiety, because the phobic object (now a conditioned stimulus) is avoided and therefore extinction cannot occur.

Avoidance is an important issue in clinical psychology. Avoidance responses are operants that prevent the occurrence of aversive consequences before they are actually experienced. This behavior is maintained by negative reinforcement. Clinically, the liability of avoidance behavior is that the person engaging in such behavior does not experience the opportunity to test whether the anticipated aversive consequences are still in effect. Thus, the circumstances that led to the initial aversive consequences may have changed, but if the person continues to avoid the original stimulus conditions, the changes will go undetected. There may also be avoidance of other stimuli due to generalization that leads to additional restrictions in healthy functioning.

Several clinical interventions address such problems. Treatments for phobias involve therapeutic interventions that prevent or remove the instrumental benefits of avoidance. Phobia treatment involves a classical conditioning paradigm in which the behavior therapist uses exposure to the conditioned stimulus to bring about extinction (see Behavior Therapy: Psychological Perspectives). The key to successful treatment is the prevention of avoidance, which would negatively reinforce the phobic behavior (Barlow 1990).

Another clinical problem that is treated, in part, by preventing avoidance behavior is obsessive-compulsive disorder (OCD). In OCD, the client experiences intrusive thoughts or images that produce anxiety. For example, someone might be obsessed with a concern that they have failed to lock their house adequately. The thoughts are high in frequency, do not feel natural to the client, and are not under the voluntary control of the client. Obsessions are thoughts or images. They are often accompanied by compulsive behaviors that serve to reduce the obsessive thoughts. In this example, a client may go back repeatedly to check that the front door is locked, preventing them from going to work. The psychological intervention used to treat OCD is exposure to the situation that produces the obsessive behavior and response prevention so that the compulsive behaviors are not emitted (Foa et al. 1980). Eventually, the anxiety associated with the problematic stimulus extinguishes, because the function of acting to reduce the distress extinguishes.

3.2 Classical Conditioning

Classical conditioning is recognized as the simplest form of associative learning. An association between a signaling stimulus (conditioned stimulus or CS) and a response producing stimulus (unconditioned stimulus or UCS) forms when the CS is presented shortly before UCS onset. The CS gradually comes to elicit a response (CR) similar to that evoked initially by the UCS. A considerable body of research beginning in the 1930s (see Patterson 1976) attempted to demonstrate that spinal reflex circuits show the associational learning of classical conditioning. While beset with theoretical and methodological difficulties, the evidence supported the ability of spinal circuits to support long-lasting (days) changes due to temporal association. Other data (e.g., Beggs et al. 1985) indicate that classical conditioning procedures produce a variety of long-term neural alterations closely approximating associative learning in the intact animal. There is some suggestion that the ability of the spinal cord to sustain this neural plasticity decreases for several days after spinal transection, but may return within a few weeks, presumably after neural reorganization.

3 Learning Conditioned Responses

In studies with animals, optimal learning is usually reached in about 130 trials when each trial has the onset of the tone conditioned stimulus preceding the onset of the air puff unconditioned stimulus by 250 ms. The interval between trials might average 30s. Intertrial intervals of 9s or less do not support conditioning. In studies with humans, optimal learning is usually reached in about 50 trials when each trial has the onset of the tone conditioned stimulus preceding the onset of the air puff unconditioned stimulus by 500 ms. The interval between trials might average 60s. Humans must additionally be distracted during conditioning, for example, by watching silent movies. Recent studies show, however, that humans who learn can accurately describe the stimulus contingencies.

Learned responses are defined as eye blinks that occur after the onset of the tone conditioned stimulus and before the air puff unconditioned stimulus on paired trials. A further criterion is that the response must be 0.5 mm or larger to be counted as a response. On test trials where only the tone conditioned stimulus is given, conditioned responses are defined as any 0.5 mm or larger response after the tone until the end of the trial, usually a period of 250 ms after the air puff would have begun. In both cases, reflexive alpha responses that occur too quickly after the tone conditioned stimulus should be excluded as not being true conditioned responses. Typically for rabbits any response latency less than 25 ms is considered to be an alpha response. For humans, any response latency less than 100 ms might be considered to be an alpha response. A common criterion for learning is the first time that eight conditioned responses occur in nine trials (89 percent responding). An additional criterion for stable performance (overtraining) requires continued performance at some level, for example 70 percent responding with learned responses. With rabbits 100 percent responding is often achieved. With humans the rate is usually not perfect. Once learned, the association shows good retention over months and perhaps years.

Traditionally, the unconditioned response was thought to be stable and impervious to influence by the training stimuli. However, reflex facilitation is seen before conditioned responses appear. Reflex facilitation occurs when the size of the reflexive response to the air puff unconditioned stimulus increases in the presence of the tone conditioned stimulus. It has also been shown that there is sensitization of the air puff unconditioned response. This sensitization is seen as an increase in the size of the air puff unconditioned response after repeated exposure to tone conditioned stimuli and air puff unconditioned stimuli that have never been presented together.

3 Conditioned Fear and the Development of Somatic CRs

According to one view (Weinberger et al. 1984), conditioned fear responses are thought to facilitate the development of somatic CRs that attenuate or totally nullify the impact of the US. The nonspecific autonomic CRs associated with conditioned fear are thought to reflect a general change in behavioral state. It not only leads to changes in activity in the autonomic nervous system, it alters the organism's orientation towards the environment in a way that enhances the reliability and efficiency of feature extraction from sensory input (Sokolov 1963), thereby facilitating the detection of stimuli that are most relevant to the development of adaptive somatic CRs. For example, conditioned decreases in heart rate (bradycardia) to a tonal CS are observed well in advance of the development of eye blink CRs to the same CS. The bradycardia CR is thought to be one of many nonspecific autonomic CRs that emerge as a result of the CS-US pairings. The altered CNS state (i.e., conditioned fear) that underlies the constellation of nonspecific autonomic CRs is thought to facilitate the development of the eye blink response. Once this adaptive somatic CR develops, the previously conditioned bradycardia disappears (Powell et al. 1990). Within this context, conditioned fear may be considered a CR that can be involved in predictive homeostasis because it motivates behavioral changes that lead to avoidance, in advance, of the presentation of an aversive stimulus that would disrupt homeostasis. The CNS changes associated with conditioned fear serve to improve the organism's ability to detect sensory stimuli that may be linked to danger and threat or may guide coping responses such as instrumentally conditioned responses that allow the organism to avoid an aversive stimulus.

Alternatively, the autonomic CRs in aversive conditioning studies may be interpreted solely in terms of autonomic responses involved in predictive homeostasis (Schneiderman 1972). The UR to an electric shock US in the restrained rabbit, for instance, involves an increase in arterial blood pressure, whereas the CR consists of bradycardia. Thus, the CR appears to mitigate the blood pressure increase elicited by the US, which becomes attenuated after several conditioning trials. In contrast, the UR to an aversive US in the unrestrained rat consists of an increase in arterial pressure as does the CR. Thus, the blood pressure CR appears to be facilitating the metabolic needs of the organism (i.e., predictive homeostasis) by what Obrist and Webb (1967) referred to as the cardiac–somatic linkage. In conclusion, it appears that much of autonomic classical conditioning can be interpreted within the context of predictive homeostasis although its relationship to such concepts as conditioned fear remain to be explored.

3.2 Classical Conditioning

A paradigm in which stimulus-elicited responses are studied that has been of continuing interest to psychophysiologists is classical conditioning. In paradigms in which the conditioned stimulus (CS) is several seconds long, with the unconditioned stimulus (UCS) occurring at CS offset, the form of the conditioned response (CR) is generally observed to be that of a heightened orienting response to the CS (Hugdahl 1995). A general question frequently investigated has concerned the relationship between classical conditioning, a very simple form of learning, and the more complex forms of learning of which humans are capable, including verbally mediated processes. For instance, Dawson and his colleagues (Dawson and Schell 1985) studied the relationship between acquisition of the conditioned skin conductance response and the human subject's conscious awareness of the relationship between the CS and the UCS, asking whether a person who is unaware of the CS–UCS contingency will show conditioning. In a series of studies in which awareness of the CS–UCS relation was prevented or delayed by a distracting secondary task, they found that a CR was seen only in subjects who became aware of the CS–UCS relation. Moreover, CRs developed only at or after the point in time in a series of CS–UCS trials when subjects indicated awareness. Thus, these results indicate that conscious relational learning is necessary for human classical conditioning of ANS responses with neutral CSs. On the other hand, the relationship between awareness and conditioning may be different with certain paradigms used to establish the skeletal eyeblink conditioned response (Clark and Squire 1999).

In most psychophysiological studies of classical conditioning, the CSs have been fairly neutral stimuli such as simple tones or colored lights. In contrast, Öhman and his colleagues (see Hugdahl 1995) conducted a series of interesting studies of a different class of CSs, those variously characterized as potentially phobic, fear-relevant, or biologically prepared, such as pictures of spiders, snakes, or angry faces. Öhman and his colleagues have found that CRs (usually the conditioned skin conductance response) conditioned to potentially phobic stimuli (a picture of a snake) are harder to extinguish than are CRs conditioned to neutral stimuli (a picture of a flower). They require more non-reinforced presentations of the CS, and are resistant to instructional manipulation. That is, while responses conditioned to neutral stimuli can usually be abolished by instructing the subject that the UCS will no longer be delivered, so that the subject no longer has any cognitive expectancy that the UCS will follow the CS, responses conditioned to potentially phobic stimuli remain after such instructions, even in the absence of conscious expectancy of the UCS. Thus, psychophysiological techniques can be used in the laboratory to study the dynamics of ‘irrational’ conditioned responses, such as one may encounter in clinical phobias where the patient fears an object while realizing rationally that the object is harmless.