NEURAL SUBSTRATES OF AVOIDANCE LEARNING

People and animals learn to avoid pain provided that warning stimuli are available to signal pain-inducing events. Such learning is generally of two types, active and inhibitory. Active avoidance refers to movements learned in response to warning stimuli for the purpose of avoiding pain. Inhibitory avoidance refers to inaction, learned because action in the presence of the warning stimuli has previously led to pain.

Limbic and Motor Systems

Research implicates the brain's limbic and motor systems in the mediation of avoidance learning. The limbic system is a vast network of interconnected regions including the amygdala, hippocampus, limbic thalamus and the cingulate area of the cerebral cortex. Relevant parts of the motor system include the striatum and the nucleus accumbens. Many laboratory studies of the neural substrates of avoidance learning involve locomotion (or its inhibition). Therefore, areas of the brain concerned with the initiation and maintenance of locomotion are also involved.

Theoretical Overview: WHAT and WHEN

Available data indicate that in avoidance learning the motor system acts as a WHAT system, and the limbic system acts as a WHEN system. The WHAT system determines what is to be done, that is, the particular behavior to be performed. Its functions include learning and remembering the response to be performed; making ready or priming the response when the avoidance situation is encountered; and executing the response. The WHAT system is relatively poor when it comes to remembering important signals in the environment, including the warning stimuli that call for avoidance behavior; the WHEN system is specialized to handle these functions. The WHEN system learns about and remembers warning stimuli, and issues command volleys of neuronal activity that tell the WHAT system precisely when to execute the avoidance response (see Figure 1).

Data supporting these ideas come from studies of the effects of experimentally induced brain damage (lesions) and from studies of the activity of brain neurons during avoidance learning. Gabriel (1993) reviewed much of the research data summarized below; citations of studies not included in the review are noted in the text that follows.

Active Avoidance Learning

Experimental lesions in the medial dorsal (MD) and anterior (AN) nuclei of the limbic thalamus, or lesions of the cingulate cortical areas that receive input from these thalamic nuclei, render rats, cats and rabbits incapable of active avoidance learning. Lesions in only one of the nuclei, or in the cingulate cortical projection field of a single nucleus, yield partial learning deficits. In addition, lesions of the amygdala block learning. In these studies laboratory tasks involve animals learning to jump over a barrier or learning to step in an activity wheel to avoid a mild electric shock signaled by tone or light warning stimuli. The deficit in animals with these lesions is a true learning deficit, not an inability to move or to perceive the warning stimuli.

The specific involvement of the limbic areas in avoidance learning receives further support from data indicating that cerebellar lesions, which block classical conditioning of eyeblink responses, do not affect avoidance learning in rabbits.

Studies of neuronal activity during active avoidance learning by rabbits in the activity wheel task have shown that amygdalar, limbic thalamic, and cingulate cortical neurons learn to produce impulses in response to the warning tone that signals shock. Impulse rate in trained rabbits increased just after the warning stimulus and reached its maximum rate just before the rabbits stepped, suggesting that stepping was triggered by the neuronal activity. That this activity is truly related to learning and not to generalized arousal is indicated by the fact that the activity is selective, for example, greater in response to the warning tone than to a second tone which is presented as often as the warning stimulus but is not predictive of shock.

Cingulate cortical neurons send axons to striatal motor areas such as the caudate nucleus. Active avoidance learning is impaired in animals with lesions in the caudate nucleus. Thus, the flow of nerve impulses from the cingulate cortex to the caudate nucleus and possibly to other motor areas is likely responsible for initiating active avoidance responses. Theoretically, the role of the caudate nucleus represents the function of the WHAT system, and the information flow from the limbic thalamus and the cingulate cortex to the caudate nucleus represents the command volley issued by the WHEN system.

During the acquisition of the stepping avoidance response by rabbits, two forms of neuronal learning have been noted in the limbic thalamus and in the cingulate cortex. One form is discriminative or selective neuronal activity (SA), as defined above, and the second form, excitatory modulation (EM), is a dramatic increase of impulse firing rate in trained rabbits in response to both warning and nonwarning tones. Two facts indicate that EM is, like SA, clearly a learning-related change and not merely a reflection of general arousal: 1. EM does not occur when rabbits experience repeated sessions with shock and nonwarning tones, that is, tones that do not predict the occurrence of shock; and 2. EM occurs at different rates in different nuclei of the limbic thalamus. (General arousal or excitement would be expected to increase neuronal activity simultaneously in many brain areas.) Although they are combined in the limbic thalamus, EM and SA have different brain origins.

Origins of EM

Experimental lesions in the large fiber tract (the mammillothalamic tract) that runs from the hypothalamus to the AN block the development of EM in the AN; these lesions diminish performance efficiency of the avoidance response. Binding of the high-affinity ligand oxotremorine to muscarinic acetylcholine receptors increases in the AN during learning (Vogt et al., 1991), and scopolamine hydrobromide, which blocks these receptors, abolishes both the EM in the AN, as well as performance of the avoidance behavior. These results suggest that EM is due to stimulation by acetylcholine of the increased numbers of receptors in the AN. Acetylcholine is released by the terminals of brain stem tegmental neurons that project to the AN. Cholinergic stimulation may increase the excitation of AN neurons by enhancing the release of the mammillothalamic tract neurotransmitter or by increasing the excitability of AN neurons in response to that neurotransmitter. To summarize, EM originates in the limbic thalamus and has the function of amplifying the selective discharges as they are relayed to the cingulate cortex.

Selective Neuronal Activity in Two Functional Circuits

Immediately after the onset of training, SA in the anterior cingulate cortex develops very rapidly, whereas neurons of the posterior cingulate cortex and in the AN develop selective activity gradually. Separate lesions in these areas induced before training impair respectively, performance in the early and late stages of learning. These two circuits represent respectively the operations of an immediate or primary working memory circuit and a secondary or intermediate-term memory circuit. Development of SA in the secondary circuit is a result of short-term consolidation processes. SA also develops very rapidly in the parts of the ascending sensory (auditory) pathway for hearing, including the medial geniculate nucleus (MGN). Because MGN neurons project to the amygdala, it was believed that SA in the MGN engenders the SA in the amygdala. However, SA in the MGN was blocked during training of subjects with temporary amygdalar lesions, indicating that the amygdala engenders SA in the MGN.

Origins of Selective Activity

Additional studies have shown that the amygdala modulates the MGN by way of an amygdalar pathway to the auditory cortex and from there to the MGN. Auditory cortical lesions abolish the rapid SA in all areas, including the MGN, basolateral nucleus of the amygdala and the anterior cingulate cortex (Duvel et al., 2001). Subjects with these lesions make fewer learned responses than controls during the early sessions of discriminative avoidance learning, and they fail to discriminate between the CS+ and the CD-during the early sessions. However, the slower SA of the secondary circuit remains intact and these subjects eventually learn to normative levels of performance. Large permanent lesions of the amygdala block the development of gradual SA in the secondary memory circuit and neurons of the lateral anterior nucleus of the amygdala exhibited slow SA. These results indicate that 1. the amygdala initiates SA in the MGN and possibly in the auditory cortex; 2. the auditory cortex is the origin of SA in the remaining areas of the primary memory circuit; and 3. the lateral anterior area of the amygdala may initiate gradual SA of the secondary memory circuit.

Inhibitory Avoidance Learning

Experimental procedures used to study inhibitory avoidance learning include the delivery of shock after performance of an unlearned response by rats (moving from a lighted area to an innately preferred dark area), the delivery of shock after performance of a previously learned response (running in a maze alley for food or water reinforcement), or assessing choice behavior such as the relative amount of time spent in neutral areas compared with time spent in areas previously established as dangerous. The timing of avoidance responses is not greatly critical for learning of these tasks because subjects are not required, as they are in active avoidance tasks, to produce discrete behaviors at particular moments. Instead, behavior must be suppressed in response to configurations of static, continuously present environmental stimuli.

The hippocampus, a limbic area of the cerebral cortex, is important to complex memory functions. Hippocampal lesions are detrimental to inhibitory avoidance learning. Hippocampal involvement may reflect the fact that the warning stimuli in inhibitory avoidance tasks are often experimental environments or places, rather than discrete stimuli such as tones. Successful performance in such tasks depends on the cognitive mapping functions of the hippocampus, including remembering whether particular environments are dangerous or safe (see Nadel, O'Keefe, and Black, 1975).

As in active avoidance learning, outputs of the amygdala are be involved in initiating learning-related plasticity in other areas in response to aversive stimulation received during inhibitory avoidance training. Research of James L. McGaugh (2000) indicates that amygdalar outputs induce the hippocampus to store or consolidate memory underlying inhibitory avoidance learning. Hippocampal cognitive mapping operations could give rise to the suppression of movement toward dangerous environments as a result of information flow over a massive WHEN system pathway from the hippocampal formation to the nucleus accumbens. The latter area is a WHAT system component that has been implicated in the suppression of locomotion.

Finally, substantial evidence indicates that learning of immobility in response to discrete cues using Pavlovian training procedures appears to be mediated by plasticity that develops within the amygdala. In this instance, the amygdala is viewed as a site of memory storage rather than a plasticity-initiating agent.

See also:ACTIVE AND PASSIVE AVOIDANCE LEARNING: BEHAVIORAL PHENOMENA; GUIDE TO THE ANATOMY OF THE BRAIN: AMYGDALA; NEURAL SUBSTRATES OF CLASSICAL CONDITIONING: FEAR CONDITIONING, FREEZING; PASSIVE (INHIBITORY) AVOIDANCE, FEAR LEARNING

Bibliography

Duvel, A. D., Smith, D. M., Talk, A., and Gabriel, M. (2001). Medial geniculate, amygdalar and cingulate cortical training-induced neuronal activity during discriminative avoidance learning in rabbits with auditory cortical lesions. Journal of Neuroscience 27, 3,271-3,281.

Gabriel, M. (1993). Discriminative avoidance learning: A model system. In M. Gabriel and B. Vogt, eds. Neurobiology of cingulate cortex and limbic thalamus, 478-523. Toronto: Birkhauser.

McGaugh, J. L. (2000). Memory—a century of consolidation. Science 287, 248-251.

Nadel, L., O'Keefe, J., and Black, A. (1975). Slam on the brakes: A critique of Altman. Brunner and Bayer's response-inhibition model of hippocampal function. Behavioral Biology 14, 151-162.

Vogt, B. A., Gabriel, M., Vogt, L. J., Poremba, A., Jensen, E. L., Kubota, Y., and Kang, E. (1991). Muscarinic receptor binding increases in anterior thalamus and cingulate cortex during discriminative avoidance learning. Journal of Neuroscience 11, 1,508-1,514.

MichaelGabriel