Neural Basis Of Classical Conditioning Research Paper

1. Introduction And Historical Background

Classical or Pavlovian conditioning is the simplest form of associative learning by which animals, including humans, learn relations among events in the world so that their future behaviors are better adapted to their environments (Rescorla 1988). Generally, classical conditioning ensues when an initially neutral stimulus (conditional stimulus, CS) is paired in close temporal proximity with a biologically significant stimulus (unconditional stimulus, US) that elicits an unlearned, reflexive behavior (unconditional response, UR). Through CS–US association formation, the animal learns to exhibit a learned behavior (conditional response, CR) to the CS that generally (a) resembles the UR (but not always), (b) precedes the US in time, and (c) reaches a maximum at about the time of US onset. A typical classical conditioning arrangement is represented in Fig. 1. [It should be noted that the contingent (informational), rather than contiguous (temporal), relationship between CS and US is essential in classical conditioning. For detailed treatment of this see Kamin 1968, Rescorla, 1968, Wagner et al. 1968.]

Classical conditioning was first characterized by Ivan P. Pavlov, a Russian physiologist. [The English translation of Pavlov’s book Conditioned Reflexes was first published in 1927. Classical conditioning was also independently discovered by an American Psychologist, Edwin B. Twitmyer.] Having already conducted prominent work on the digestive system, for which he received a Nobel Prize in 1904, Pavlov employed a salivary-conditioning procedure in dogs to systematically characterize some of the fundamental principles of classical conditioning. In brief, Pavlov’s dogs were presented with a discrete CS (e.g., the beat of a metronome) just prior to the delivery of a US (e.g., meat powder). Initially, the subjects did not respond to the CS but salivated profusely (UR) to the US. With repeated CS–US pairings, Pavlov’s dogs exhibited salivation (CR) to the CS that both preceded US onset and occurred in the absence of the US.

2. Behavioral Principles Of Classical Conditioning

Since Pavlov, various types of classical conditioning procedures have been developed, ranging from a potently fast (one-trial) taste aversion conditioning (e.g., Garcia et al. 1974) to a relatively slow and incremental eyeblink conditioning (e.g., Gormezano et al. 1983), and a wide variety of organisms have been used, ranging from invertebrates (e.g., Aplysia) to primates (e.g., humans). These different types of classical conditioning, however, all share certain common factors that influence the formation of CS–US associations which can be classified into three general categories: contiguity (or temporal) constraints, sensory constraints, and contingency (or informational) constraints (Table 1). Clearly, these constraints must be considered when employing classical conditioning in learning and memory research.

In addition to simple CS–US pairings (also called first order conditioning), there are many other training protocols within classical conditioning that can serve as effective tools for investigating the theoretical and biological mechanisms of learning and memory. Some of these are presented in Fig. 2.

3. Classical Conditioning As A Model System For Studying The Neurobiology Of Learning And Memory

Current views recognize that, in mammals, there are multiple forms or aspects of learning and memory (e.g., habituation, sensitization, classical conditioning, priming, procedural learning, episodic, semantic) that are subserved by different structures in the central nervous system (e.g., reflex pathways, cerebellum, amygdala, striatum, diencephalon, medial temporal lobe, neocortex) (Squire 1987). Considerable progress has been made in the understanding of these different learning and memory systems through the use of various experimental techniques (e.g., lesions, reversible inactivation, drug administration, neural recordings, genetic manipulations, brain imaging). The success of this work is due in large part to the use of classical conditioning as a model system, since it provides an important advantage over other, more complex, forms of learning in that the stimuli involved (CS and US) are well defined and can be precisely controlled, and the behavioral output is discrete and may be accurately assessed.

3.1 Localization Of Brain Substrates Of Classical Conditioning: Rationale

The identification of the locus of learning and memory storage in the brain is a prerequisite to an understanding of the neurochemical, cellular and molecular mechanisms by which organisms acquire and retain information. For many years the task of localizing memory storage (the engram) has been the paramount challenge facing investigators of memory mechanisms in mammalian systems. Classical conditioning is especially attractive as a model system for use in these types of investigations since only two stimuli are involved, and thus the learning or association of CS and US must occur at the brain site(s) where the two pieces of information converge. For instance, in classical eyeblink conditioning, where animals (such as mice, rats, rabbits or humans) learn to exhibit eyeblink responses to a CS (e.g., tone) that has been paired with a US (e.g., air-puff to the eye), one can trace the pathways from the peripheral sensory receptors in the ear (for CS) and around the eye (for US) to the brain and examine those brain regions where the CS and US pathways converge. It is only in the past 20 years, however, that technology has permitted this type of analysis, and only then at the gross structural level.

The convergence of CS and US information, however, is not a sufficient condition for identifying the locus of learning. In order for a particular region of the brain to be considered a viable candidate as a learning and memory site, it must demonstrate the following criteria:

(a) permanent lesions of the putative site prior to CS–US training should completely and permanently abolish the acquisition of the CR;

(b) permanent lesions made after training should completely abolish the expression of the CR, and the CR should not be reacquired with further CS–US pairings;

(c) reversible inactivation during CS–US training should block the development of the CR such that when the structure is activated again the CR should develop comparably to that of a naıve animal (i.e., no evidence of savings);

(d) reversible inactivation following training should temporarily impair the expression of the CR;

(e) learning-related neural activities should occur that correspond with and immediately precede the behavioral CR;

(f) electrical stimulation of the CS and the US input pathways to the putative site should effectively substitute for the peripheral stimuli and support conditioning; and

(g) electrical stimulation of the putative site should evoke the CR.

If there is a single locus of learning that supports classical conditioning, then these seven criteria must be demonstrable within that specific learning site; whereas if there are multiple structures that encode CS–US association, then these seven criteria must be satisfied collectively by these structures. Figure 3 illustrates how these different criteria can be applied to a putative locus of CS–US association.

In Fig. 3, the afferent CS and US information is relayed via hypothetical structures 1 (and 5 for different CSs) and 2 (for US), respectively, to the learning site 3. The outputs (efferents) from structure 3 activate the motor center 4 that in turn controls the CR. Permanent lesions of structures 1, 2, 3 or 4 prior to conditioning will block the acquisition and or expression of the CR. Thus, the permanent lesion technique (e.g., electrical, chemical, radio-frequency, aspiration, ischemia) is limited in that it does not allow for dissociation of the site of learning from the sites of input or from motor centers. In contrast, the reversible inactivation (pharmacological, cooling) technique, which temporarily inactivates the neurons within a structure, is not so constrained. For example, reversible inactivation of structures 1, 2 or 3 during conditioning will block acquisition of the CR to CS . In subsequent CS–US training when the inactivation has been reversed, the animal should learn as though it is naıve. By contrast, reversible inactivation of structure 4 at the time of CS–US training will block the expression of the CR, but once the inactivation is removed, the animal will immediately exhibit CRs to the CS because the CS–US association center was not affected during conditioning.

Thus structure 4 must be efferent to the site of learning. Structures 1, 2 and 3 can be further dissociated by examining the effects of reversible inactivation to CSs of different sensory modalities. Whereas inactivation of structure 2 and 3 will block conditioning to all CS modalities (e.g., CS and CS ), inactivation of structure 1 will block conditioning to CS but not CS . Thus structure 1 must be afferent to the site of learning. Finally, structures 2 and 3 can be dissociated by examining the reversible inactivation effects following conditioning: once the animal has acquired the CR, inactivating structure 3 should abolish the expression of CR, whereas inactivation of structure 2 should not interfere with the expression of the CR because the CS inputs to the CS–US association center remain intact. Interestingly, with continued CS–US training, inactivation of structure 2 will lead to extinction of the CR because the CS–US association center will, as a result, receive only the CS information ( just as in CS-alone extinction training). If structures 1 and 2 relay information about the CS and the US, then electrical stimulation of the CS and US pathways should also support conditioning. Lastly, recordings from structure 3 (the site of learning) should reveal learning-related changes in neural activity that model the behavioral CR, whereas recordings from CS and US input structures should show stimulus-evoked neural activities.

The neural circuitry of the brain is almost infinitely complex and thus no single experimental technique is in itself sufficient to identify the site of learning. It is also important to note that different types of classical conditioning (e.g., eyeblink versus fear conditioning) are subserved by different neural circuits. Through the utilization of various techniques, however, the structures and mechanisms involved in this type of learning can be reasonably delineated and systematically analyzed for its validity.

3.2 Putative Cellular Mechanisms Of Classical Conditioning

Once the learning and memory storage site has been identified, a logical next step is to determine what neural changes take place that allow a CS that previously did not elicit a CR (before conditioning) to now evoke a CR (after conditioning). It is generally assumed that an initially weak connection between a CS relaying structure and a CS–US association structure becomes strengthened as a function of CS–US paired training, such that the CS becomes able to effectively activate the CS–US association structure, which in turn activates the CR pathway. It is further hypothesized that changes in synaptic efficacy (that is, the efficacy with which a neuron is able to communicate with another neuron) underlie this type of strengthening of the CS relay-to-CS–US association connection (e.g., Hebb’s postulate (Hebb 1949)). Two forms of experimentally-induced synaptic plasticity, long-term potentiation (LTP) and long-term depression (LTD), have received close scrutiny as the most promising cellular mnemonic mechanisms (Bliss and Collingridge 1993; Ito 1989). LTP and LTD refer to sustained increase and decrease of synaptic transmission, respectively, following different stimulation patterns of afferent fibers. In brief, LTP is characterized by its rapid inducibility and longevity (lasting of the order of hours in vitro to weeks in vivo), as well as its being strengthened by repetition and demonstrating specificity and associativity. LTD displays similar characteristics desirable of an information storage mechanism. Both LTP and LTD have been demonstrated in various brain structures, including those that are hypothesized to be critical for learning and memory.

One can easily imagine how LTP and LTD might be applicable to classical conditioning. For example, suppose that the CS pathway to the site of learning (pathway between steps 1–3 in Fig. 3) is initially weak and that, as a result, the CS alone cannot sufficiently activate the CS–US association center to produce a CR. Following CS–US pairings, if this pathway is strengthened via LTP (or LTP-like changes), then the CS alone will be able to activate the CS–US association center to elicit a CR. Similarly LTD or LTD-like changes may support conditioning by weakening the inputs to a structure that normally inhibits the CS–US association center (as is postulated to occur in eyeblink conditioning circuit). It is likely that combinations of LTP and LTD-like forms of synaptic plasticity (and perhaps other unknown cellular changes), involving a network of synapses, are important in classical conditioning.

3.3 Neuronal Substrates Of Eyeblink Conditioning

Classical eyeblink conditioning in rabbits has been used extensively to investigate brain mechanisms underlying learning and memory. Converging lines of evidence from lesion, recording, stimulation, reversible inactivation, and brain imaging studies indicate that the cerebellum mediates the formation of the CS–US association for eyeblink conditioning (Thompson 1990). In brief, selective lesions of the cerebellum (i.e., the interpositus nucleus) block the acquisition and retention of eyeblink CRs; the lesion is limited to the CR since the UR (reflexive eyeblink) to the US is not affected. (An important feature of eyeblink conditioning is that the CR and the UR can be dissociated and, thus, effects of various manipulations on memory versus performance can be carefully addressed.) Correspondingly, recording studies indicate that cells in specific regions of the cerebellum undergo plastic changes during eyeblink conditioning; for example, cells in the interpositus nucleus increase their activity (postulated to occur via LTP-like changes), while Purkinje cells in the cortex, which send inhibitory projections to the interpositus nucleus, decrease their activity (postulated to occur via LTDlike changes).

The involvement of the cerebellum in eyeblink conditioning is also evidenced by stimulation studies which show that direct stimulation of the two major afferents to the cerebellum, the mossy fibers from the pontine nucleus and the climbing fibers from the inferior olive, can substitute for the peripheral CS and US, respectively. Since limited lesions of the pontine nucleus (i.e., the lateral region) abolish CRs to a tone CS but not to a light CS, and lesions of the inferior olive in animals that already acquired CRs result in behavioral extinction with continued CS–USpaired training, it is not likely that these afferent structures are the site of learning and memory storage. Reversible inactivation studies further support the conclusion that the cerebellum, and not its efferent structures, is the locus of CS–US association. Moreover, recent human brain-imaging studies reveal eyeblink conditioning-related activity changes in the cerebellum. Collectively, these findings strongly suggest that the cerebellum is essential for eyeblink conditioning. Although the cerebellum seems to be critical, the relative importance of the cerebellar cortex and the interpositus nucleus in supporting eyeblink conditioning is not clear and has been disputed (see Kim and Thompson 1997 for detailed treatment of this topics and a putative eyeblink conditioning circuit). The fact that the critical CS–US association in eyeblink conditioning occurs within the cerebellum, however, permits research at the molecular level of analysis.

3.4 Neuronal Substrates Of Fear Conditioning

In fear conditioning, CSs such as tones, lights or experimental chambers are typically paired with aversive US such as electric shock. Following CS–US pairings, the CS can elicit numerous fear CRs, such as an increase in blood pressure, reduction in pain sensitivity (analgesia), fear-potentiated startle, and or defensive freezing. Several lines of evidence point to the amygdala, one of the principle structures of the limbic system that seems to be situated such that it has access to both sensory inputs and response outputs, as a critical neural substrate for this type of emotional learning (LeDoux 1996). In brief, the critical role of the amygdala in fear conditioning is supported by observations that:

(a) amygdalar lesions (permanent and reversible) abolish various fear CRs as well as innate fear responses;

(b) selective lesions of structures afferent to the amygdala affect conditioning to specific CSs (e.g., medial geniculate nucleus of the thalamus for tones, and the hippocampus for contexts);

(c) selective lesions of structures efferent to the amygdala abolish specific CRs (e.g., the lateral hypothalamus for blood pressure CR, and the ventral region of the periaqueductal gray matter for freezing CR);

(d) recording studies reveal that neurons in the amygdala respond to both CS and US and undergo plastic changes during fear conditioning (e.g., LTPlike changes);

(e) electrical stimulation of the amygdala elicits fear responses; and

(f) drugs that block LTP also prevent fear conditioning.

However, there is also evidence that the amygdala may not necessarily be involved in fear learning and memory, and that other brain structures (e.g., insular cortex) may mediate fear conditioning (McGaugh et al. 1996). Thus, additional studies are required to firmly establish the role of the amygdala in fear conditioning.

4. Conclusion

Although classical conditioning is well understood at a behavioral level and the neuroanatomical circuits that underlie it are beginning to be unveiled, there is much left to learn. However, this most basic form of associative learning offers a useful means to investigate the synaptic and molecular mechanisms underlying learning and memory and may help reveal common biological mechanisms shared by all learning and memory systems.

Bibliography:

1. Garcia J, Hankins W G, Rusiniak K W 1974 Behavioral regulation of the milieu interne in man and rat. Science 185: 824–31
2. Gormezano I, Kehoe E J, Marshall B S 1983 Twenty years of classical conditioning research with the rabbit. In: Sprague J M, Epstein A N (eds.) Progress in Psychobiology and Physiological Psychology. Academic Press, New York
3. Hebb D O 1949 The Organization of Behavior: A Neuropsychological Theory. John Wiley and Sons, New York
4. Ito M 1989 Long-term depression. Annual Review of Neuroscience 12: 85–102
5. Kamin L J 1968 Attention-like processes in classical conditioning. In: Jones M R (ed.) Miami Symposium on the Prediction of Behavior: A versive Stimulation. University of Miami Press, Miami, FL, pp. 9–32
6. Kim J J, Thompson R F 1997 Cerebellar circuits and synaptic mechanisms involved in classical eyeblink conditioning. Trends in Neurosciences 20: 177–181
7. LeDoux J 1996 The Emotional Brain. Simon vans Schuster, New York
8. McGaugh J L, Cahill L, Roozendaal B 1996 Involvement of the amygdala in memory storage: Interaction with other brain systems. Proceedings of the National Academy of Sciences of the United States of America 93: 13508–14
9. Pavlov I P 1927 Conditioned Reflexes. Oxford University Press, London
10. Rescorla R A 1968 Probability of shock in the presence and absence of CS in fear conditioning. Journal of Comparative and Physiological Psychology 66: 1–3
11. Rescorla R A 1988 Behavioral studies of Pavlovian conditioning. Annual Review of Neuroscience 11: 329–52
12. Squire L R 1987 Memory and Brain. Oxford University Press, NY
13. Thompson R F 1990 Neural mechanisms of classical conditioning in mammals. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences 329: 161–70
14. Wagner A R, Logan F A, Haberlandt K, Price T 1968 Stimulus selection in animal discrimination learning. Journal of Experimental Psychology 76: 171–80