Cerebellum And Associative Learning Research Paper

1. The Cerebellum And Classical Eyeblink Conditioning

Arguably the best-understood associative learning paradigm, from both behavioral and neurobiological perspectives, is classical conditioning of the eyeblink response. Briefly, a neutral stimulus such as a tone or light (the conditioned stimulus or CS) is presented just before an aversive stimulus such as a peri-orbital shock or corneal air puff (the unconditioned stimulus or US). Initially, the CS produces no overt movement while the US causes a reflexive eyeblink (the unconditioned response or UR). After 50–100 pairings of the CS and US, the CS begins to elicit a learned eyeblink (the conditioned response or CR). While most eyeblink conditioning experiments have involved rabbits as subjects, it appears that all mammals, including humans, learn this simple associative task at similar rates using similar brain circuitry.

For a variety of reasons that include (a) the relative simplicity of the response being monitored, (b) the great deal of control that the experimenter has over stimulus delivery, and (c) the precise timing of the learned response, this behavioral task has proven useful for delineating the neural circuitry involved in simple associative learning. Many experiments conducted since the early 1980s have demonstrated conclusively that the cerebellum contains a population of neurons that change their patterns of firing to encode the acquisition and performance of the classically conditioned eye-blink response—that is, the cerebellum’s circuitry constitutes the essential learning and memory architecture for this basic associative learning procedure (see Woodruff-Pak and Steinmetz 2000, for review).

1.1 Lesion Experiments

The initial demonstrations of the involvement of the cerebellum in classical eyeblink conditioning were lesion experiments (e.g., McCormick and Thompson 1984). Lesions placed in the interpositus nucleus of the cerebellum prevented acquisition of eyeblink CRs and abolished previously learned CRs. Lavond et al. (1985) demonstrated the same lesion effect with infusions of kainic acid, which spared fibers of passage that course through or near the interpositus nucleus. Reversible lesions placed by cooling brain tissue (Clark and Lavond 1993) or injecting muscimol (Krupa et al. 1993) were also effective in abolishing CRs. Interestingly, when additional paired training was delivered without cooling or muscimol inactivation, the animals showed no savings in the rate of CR acquisition: they behaved as if they had received no previous paired training. These studies provide strong evidence that critical neuronal plasticity that underlies classical eyeblink conditioning occurs in the cerebellum.

Lesions of the cerebellar cortex have not produced as consistent results as interpositus nucleus lesions (e.g., Lavond and Steinmetz 1989). Cerebellar cortical lesions have reportedly caused retarded rates of CR acquisition, reduced CR amplitudes, or the appearance of mistimed CRs. These data indicate that the cerebellar cortex is involved in the conditioning process, but its precise role in conditioning or its interactions with the interpositus nucleus during conditioning are not well understood.

1.2 Recording Experiments

Electrophysiological recordings taken from cerebellar cortex and the interpositus nucleus have provided additional evidence for the involvement of the cerebellum in classical eyeblink conditioning (e.g., Berthier and Moore 1986, 1990). Recordings in regions of the cerebellum known to receive converging CS and US input have revealed neurons that discharge with patterns that seem to be encoding the conditioning process. Specifically, Purkinje cells were identified that discharged when the CS or US was presented. Other Purkinje cells either increased or decreased their rate of discharge in a pattern that seemed to be time-locked to execution of the behavioral CR. Purkinje cells that decreased their firing rates are particularly interesting because Purkinje cells are known to inhibit neurons in the deep cerebellar nuclei. Thus, a decrease in firing rate of a Purkinje cell could result in an increase in excitability of interpositus nucleus neurons, a result that is compatible with formation of a behavioral CR.

Similar to cerebellar cortex, neurons that developed discharge patterns highly correlated with CR performance were observed in the interpositus nucleus. Neurons that discharged to presentations of the CS and US were seen and, after learning, neurons that discharged in patterns that were time-locked with the behavioral CR were abundant. Interestingly, the onset of interpositus unit activity preceded the behavioral response by 30–60 milliseconds. These important observations provide strong evidence that cellular activity in the interpositus nucleus is the neural substrate of the behavioral CR that is observed. It is thought that CR-related activity generated in the interpositus nucleus activated neurons in the red nucleus which, in turn, activate neurons in the brainstem motor nuclei that are responsible for generating eyeblinks.

1.3 CS And US Pathways

Using stimulation, recording and lesion methods (e.g., Steinmetz et al. 1989), the putative pathways for projecting CS and US information from the periphery to the cerebellum have been delineated. It appears that a tone CS is projected from the ear to the cochlear nuclei which, in turn, relays tone information to the basilar pontine nuclei. The pontine nuclei then project information about the CS to the cerebellum along mossy fibers. On the US side, air puffs are known to activate corneal receptors that send projections to the trigeminal nucleus. The trigeminal nucleus, in turn, projects information about the US to the rostromedial portion of the dorsal accessory inferior olive. Climbing fibers that originate from the inferior olive then relay information about the occurrence of the US to the cerebellum.

1.4 The Cerebellum As An Associator

There is ample evidence from anatomy, electrophysiology, lesion, and micro stimulation experiments that information concerning the occurrences of the CS and US converges on populations of neurons in the cerebellum. The leading models of the involvement of the cerebellum in classical eyeblink conditioning postulate that CS–US inputs converge in two locations— the interpositus nucleus and cerebellar cortex (e.g., Steinmetz 2000). Changes in the excitability of cortical and nuclear neurons, produced by convergent CS and US inputs, are thought to form the cellular bases for the learning and performance of the classically conditioned eyeblink response. In essence, the cerebellar circuitry serves as an ‘associator’ for discrete stimuli that are presented in the environment. This idea is certainly not new. Computational neurobiologists such as Marr (1969) have long considered the architecture of the cerebellum to be ideal for associating environmental information with teaching or reinforcing inputs. Further, models such as those of Marr have hypothesized that mossy fibers and climbing fibers serve as the environmental and teaching inputs, respectively. This architecture maps very nicely onto the known neural circuitry involved in eyeblink conditioning where the CS appears to be carried along mossy fibers and the US appears to be carried along climbing fibers.

The previous models do not, however, predict a role for the deep cerebellar nuclei in the associative learning process; rather they postulate that the deep nuclei are passive recipients of outflow from cerebellar cortex. The data collected using eyeblink conditioning suggest differently. The nuclei appear to receive convergent CS–US input, the nuclei show neuronal responses that are related to conditioning, and the reversible lesion experiments of Clark and Lavond (1993) and Krupa et al. (1993) suggest that critical cellular plasticity related to conditioning occurs in the nuclei. At this time, the most parsimonious explanation of the available data suggest that critical plasticity that underlies classical eyeblink conditioning occurs both in cerebellar cortex and in the deep cerebellar nuclei. It has been suggested that these two areas may encode different features of the conditioning, with excitability changes in nuclear cells important for generating activity that drives brainstem motor neurons responsible for eyeblink CRs, and excitability in cortical cells important for providing gain on the response and for regulating the timing of the response (e.g., Gould and Steinmetz 1996, Steinmetz 2000).

2. The Cerebellum And Other Associative Learning Paradigms

There are surprisingly few other demonstrations of the involvement of the cerebellum in associative learning. This is likely not due to a general lack of involvement of the structure in this type of learning (although see below for some limitations on cerebellar involvement in associative learning) but rather that few experiments have been conducted to explore the involvement of the cerebellum in associative learning tasks. Two exceptions are briefly described here; adaptation of the vestibulo–ocular reflex (VOR) and instrumental signaled bar-press conditioning.

2.1 Adaptation Of The VOR

The VOR is a brain system used to stabilize a visual image on the retina during movement. In this reflex, rotations of the head are detected by semicircular canals located in the inner ear, and the eyes are moved in their sockets in the direction opposite to the movement of the head. This reflex stabilizes the line of sight. The VOR is highly plastic as the gain of the reflex can be changed easily to accommodate changes in the strength or efficiency of the extraocular muscles to deal with changing levels of vestibular activation. In many respects, this can be considered to be an associative learning procedure as it is known that gainsetting of the VOR is dependent on two events: vestibular input from the semicircular canals and visual information (the relative slippage of the visual image on the retina during head movements) to determine if a change in VOR gain is needed. Over the years, a variety of studies have implicated the cerebellum and associated brainstem structures in adaptation of the VOR (Ito 1984, Lisberger 1988). In a similar way to classical eyeblink conditioning, it appears that neuronal plasticity that forms the basis of VOR adaptation occurs in discrete regions of the cerebellar cortex and in brainstem nuclei (the vestibular nuclei) that receive convergent input from the vestibular and visual systems.

Critical involvement of the cerebellum has also been demonstrated for an instrumental conditioning task that has some similarities to classical eyeblink conditioning (Steinmetz et al. 1993). In this task, rats are first shaped to press a bar to terminate a mild, pulsating foot-shock. After learning the bar-press response, the rats are placed in a signaled-training situation where they learn to avoid the foot-shocks by pressing the response bar during tone presentations. Rats reach 50–60 percent avoidance rates with about 10–12 days of training in this procedure. This associative task is somewhat similar to classical eyeblink conditioning in that a neutral stimulus (a tone) is used to signal an impending noxious stimulus (a mild footshock). Bilateral lesions of the dentate and interpositus nuclei in rats prevented learning of this avoidance response. Escape responding was initially high in lesioned rats, but this responding decreased over sessions. Interestingly, deep nuclear lesions seem to be effective in preventing avoidance learning only when the interval between tone onset and foot-shock onset is five seconds or less.

3. When Is The Cerebellum Critical For Simple Associative Learning?

Another way to frame this question is to ask: when is the cerebellum not involved in associative learning? A number of experiments have addressed this issue.

First, the cerebellum appears to be necessary for associative learning when the interval of time between the stimuli being associated is relatively short. Classical eyeblink conditioning can only be obtained when the CS–US interval is 3–4 seconds or less. Longer CS–US intervals do not produce eyeblink CRs, although a variety of other conditioned responses can be elicited. Adaptation of the VOR requires near-simultaneous occurrences of head rotation and retinal slip. As detailed above, signaled bar-press conditioning seems to be critically dependent on cerebellar function only when relatively short CS–US intervals are used. Avoidance conditioning is relatively easy to obtain with longer tone–foot-shock intervals, but cerebellar lesions appear to have no effect on these learned responses—suggesting that other brain areas are critical for active avoidance learning when intervals between the stimuli are relatively long. These observations are highly consistent with what is known about the role of the cerebellum in movement and posture— the cerebellum is intricately involved in making fine adjustments to ongoing movements during relatively brief periods of time (often referred to as movement error-correction).

Second, and related to the first point, the cerebellum seems to be critical for associative learning that involves relatively simple, discrete skeletal muscle responses. Conditioned eyeblinks, VOR adaptation, and conditioned bar-press responses require the rapid recruitment of relatively few muscles, especially when the time period allowed for responding is relatively brief. The idea that cerebellar involvement in learning may be limited by response requirements has been tested (Steinmetz et al. 1991). Rabbits were trained in two tasks: classical eyeblink conditioning, and a discriminative avoidance procedure. In the discriminative avoidance procedure, rabbits were presented with a tone and required to locomote in an activity wheel to avoid a foot-shock that was presented after the tone onset. Bilateral lesions of the interpositus nucleus of the cerebellum prevented acquisition of the conditioned eyeblink response and abolished already learned eyeblink CRs, but had no effect on the acquisition or performance of the discriminative avoidance response. While the discriminative wheelturn avoidance task differs from eyeblink conditioning on several dimensions (e.g., it involves discrimination learning and uses a longer interstimulus interval), one of the major differences between paradigms lies in the response requirements (discrete eyeblink vs. a relatively complex, bipedal, locomotive response). Lavond and colleagues (1985) have also shown that the conditioned changes in heart-rate that are normally observed during classical eyeblink conditioning are not affected by cerebellar lesions. These data suggest that the encoding of autonomic responses during associative learning involves areas outside of the cerebellum, a finding that is compatible with a number of other studies. Together, these data suggest that the cerebellum may be involved in associative learning when relatively discrete skeletal muscle responses are conditioned.

Third, there is evidence that the cerebellum is not involved in encoding associative reward learning (i.e., learning that involves reinforcement). The rats described above that showed severe deficits in learning to avoid the signaled foot-shock after cerebellar lesions (Steinmetz et al. 1993) were trained in an appetitive version of the task. In the appetitive version of the task, a tone was presented for a 3–5 second period, and a bar press resulted in the delivery of food pellet reward. The cerebellar-lesioned rats easily learned the appetitive task even though they showed a complete inability to learn the aversively motivated task. In a more direct comparison of appetitive and aversive classical conditioning, Gibbs (1992) trained rabbits on two classical conditioning tasks before delivering lesions to the interpositus nucleus of the cerebellum. One task was standard classical eyeblink conditioning (an aversive task), while the second task was classical jaw-movement conditioning (an appetitive task). Jaw movement conditioning involves pairing a tone CS with the delivery of water or juice (the US) into the mouth. The water or juice causes a movement of the jaw as the rabbit consumes the liquid. Several pairings of the CS and US produce an anticipatory jaw movement to the tone (the CR). Gibbs showed that in this within-subject experiment, lesions of the cerebellar deep nuclei abolished eyeblink conditioning but had no effect on jaw-movement conditioning. These data suggest that the cerebellum is involved in encoding aversively motivated associative learning but not appetitively motivated associative learning. This suggestion is not surprising given the large body of research that has detailed the involvement of forebrain structures and circuits in reward learning.

4. The Cerebellum And Associative Learning

For several years, theorists who have speculated about the function of the cerebellum have noted that the basic anatomy and architecture of the cerebellum seems to be designed for associative learning. The cerebellum receives inputs through two separate and unique systems of fibers: mossy fibers and climbing fibers, and a growing body of evidence suggests that plasticity in cerebellar neurons may be due to associative interactions that occur between these inputs. Further, classical eyeblink conditioning seems to be an ideal associative learning paradigm for studying the involvement of the cerebellum in associative learning. This procedure involves the conditioning of discrete responses, involves the presentation of an aversive or noxious US, and involves a relatively brief period of time between the presentation of the stimuli being associated. In essence, the classical eyeblink conditioning procedure could be considered the prototypical learning paradigm for engaging the cerebellum during associative learning. Further studies into cellular and systems-level cerebellar processes engaged during classical eyeblink conditioning should provide valuable data concerning the general role of the cerebellum in associating two or more external stimuli or internal events in time.

Bibliography:

1. Berthier N E, Moore J W 1986 Cerebellar Purkinje cell activity related to the classically conditioned nictitating membrane response. Experimental Brain Research 63: 341–50
2. Berthier N E, Moore J W 1990 Activity of deep cerebellar nuclear cells during classical conditioning of nictitating membrane extension in rabbits. Experimental Brain Research 83: 44–54
3. Clark R E, Lavond D G 1993 Reversible lesions of the red nucleus during acquisition and retention of a classically conditioned behavior in rabbits. Behavioral Neuroscience 107: 264–70
4. Gibbs C M 1992 Divergent effects of deep cerebellar lesions on two different conditioned somatomotor responses in rabbits. Brain Research 585: 395–9
5. Gould T J, Steinmetz J E 1996 Changes in rabbit cerebellar cortical and interpositus nucleus activity during acquisition, extinction and backward classical conditioning. Neurobiology of Learning and Memory 65: 17–34
6. Ito M 1984 The Cerebellum and Neural Control. Raven Press, New York
7. Krupa D J, Thompson J K, Thompson R F 1993 Localization of a memory trace in the mammalian brain. Science 260: 989–91
8. Lavond D G, Hembree T L, Thompson R F 1985 Effect of kainic acid lesions of the cerebellar interpositus nucleus on eyelid conditioning in the rabbit. Brain Research 326: 179–82
9. Lavond D G, Lincoln J S, McCormick D A, Thompson R F 1984 Effect of bilateral lesions of the dentate and interpositus cerebellar nuclei on conditioning of heart-rate and nictitating membrane eyelid responses in the rabbit. Brain Research 305: 323–30
10. Lavond D G, Steinmetz J E 1989 Acquisition of classical conditioning without cerebellar cortex. Behavioral Brain Research 33: 113–64
11. Lisberger S G 1988 The neural basis of learning of simple motor skills. Science 242: 728–35
12. Marr D 1969 A theory of cerebellar cortex. Journal of Physiology 202: 437–70
13. McCormick D A, Thompson R F 1984 Cerebellum: essential involvement in the classically conditioned eyelid response. Science 223: 296–9
14. Steinmetz J E 2000 Brain substrates of classical eye blink conditioning: A highly localized but also distributed system. Behavioral Brain Research 110: 13–24
15. Steinmetz J E, Lavond D G, Thompson R F 1989 Classical conditioning in rabbits using pontine nucleus stimulation as a conditioned stimulus and inferior olive stimulation as an unconditioned stimulus. Synapse 3: 225–33
16. Steinmetz J E, Logue S F, Miller D P 1993 Using signaled barpressing tasks to study the neural substrates of appetitive and aversive learning in rats: Behavioral manipulations and cerebellar lesions. Behavioral Neuroscience 107: 941–54
17. Steinmetz J E, Sears L L, Gabriel M, Kubota Y, Poremba A 1991 Cerebellar interpositus nucleus lesions disrupt classical nictitating membrane conditioning but not discriminative avoidance learning in rabbits. Behavioral Brain Research 45: 71–80
18. Woodruff-Pak D S, Steinmetz J E (eds.) 2000 Eyeblink Classical Conditioning: Animal Models. Kluwer, Boston