The ability of Aplysia and other gastropod molluscs to exhibit complex behaviors that can be modified by associative learning has encouraged us to search for an elementary behavior controlled by a simple and well analyzed neural circuit that also can be modified by this type of learning. Toward that end, we have now produced classical conditioning in the defensive siphon and gill withdrawal reflex of Aplysia. We used as a conditioned stimulus (CS) a light tactile stimulus to the siphon, which produces weak siphon and gill withdrawal. As the unconditioned stimulus (US), we used a strong electric shock to the tail, which produces a massive withdrawal reflex. Specific temporal pairing of the CS and US endowed the .CS with the ability of triggering enhanced withdrawal of both the siphon and the gill. Random or unpaired presentations of the CS and US, as well as presentations of the CS or US alone, produced either no enhancement or significantly less enhancement than paired presentations of the CS and US. The conditioning is acquired rapidly (within 15 trials) and is retained for several days. The conditioned response is abolished completely by removal of the abdominal ganglion and many of the neurons involved in the conditioning have been identified in this ganglion previously. These include the sensory neurons and several interneurons in the CS pathway and the siphon and gill motor neurons of the conditioned and unconditioned response pathways. Moreover, the sensory neurons of the US pathway have been identified in the pleural ganglia. As a result of its simplicity, it should be possible in this reflex to specify neurons that are causally related to the conditioned response. Since this reflex also exhibits nonassociative learning, it also may be possible to compare associative and nonassociative learning on a mechanistic level.

Brief, noxious, electrical or mechanical stimulation of the skin of Aplysia produces enhancement of defensive reflexes triggered at the same site for at least a week after the noxious stimulation. This site- specific behavioral sensitization can be expressed as an increase in duration of the siphon-withdrawal reflex and as an increase in magnitude of the tail-withdrawal reflex. It is unlikely that peripheral factors play a predominant role in the long-term memory. First, long- term enhancement is blocked when the CNS is disconnected from the noxious stimulation site by nerve transection. Second, long-term enhancement is blocked by preventing neural activation at the noxious stimulation site, indicating that persistent physical damage alone is insufficient to cause the enhancement. A role for activity-dependent extrinsic modulation (ADEM) of mechanosensory neurons is suggested by similar site-specific enhancement produced when weak sensory activation is paired with general modulation elicited by strong stimulation of a distant site. Because this pairing represents a form of classical conditioning, site-specific sensitization and cutaneous classical conditioning appear to be closely related in this system. These findings suggest that site-specific sensitization reflects, at least in part, a central, long-term memory of injury. This form of memory may be phylogenetically widespread, and functionally similar to aspects of hyperalgesia. In addition, the close relationship between site-specific sensitization and cutaneous classical conditioning supports the hypothesis that some forms of classical conditioning evolved from mechanisms of sensitization.

We thank A. Clatworthy, M. Dulin, and P. Illich for comments on an earlier version of this article, J. Pastore for preparing the figures, and N. Karin for use of cell culture facilities.

1. Cutaneous stimulation of opposite ends of the body causes qualitatively different siphon responses: tail stimulation causes flaring and backward bending (the siphon T response), whereas head stimulation causes constriction and slight anterior bending (the siphon H response). This paper characterizes the motor neuronal control of siphon T and siphon H responses. 2. The siphon response to tail nerve (p9) shock in a semi-intact preparation was indistinguishable from the siphon T response in intact or parapodectomized animals. Similarly, the siphon response to head nerve (c2) shock in this preparation was indistinguishable from the siphon H response in intact or parapodectomized animals. 3. Central siphon motor neurons (SMNs) were found to cause a wider variety of movements than previously reported. The movements produced by the LFSB cells strongly resemble the flaring response of the siphon to tail or tail nerve stimulation. The movements produced by RDS and LDS1 resemble components of the constricting response of the siphon to head or head nerve stimulation. 4. Among central SMNs, the LFSB cells show the strongest activation by posterior stimulation, whereas RDS and LDS1 show the strongest activation by anterior stimulation. The LFSA cells, which produce much weaker siphon constriction, are only activated slightly by posterior stimulation and are inhibited by anterior stimulation. Peripheral SMNs are inhibited by stimulation of head and tail nerves, and thus their activity does not directly contribute to siphon T and H responses. 5. Artificially activating central SMNs with the pattern of activity previously exhibited after tail or head nerve stimulation indicated the sufficiency of the LFSB cells for the siphon T response, and of RDS and LDS1 for the siphon H response. 6. Dramatic behavioral deficits produced by hyperpolarizing the LFSB cells during tail nerve stimulation, or by hyperpolarizing RDS and LDS1 during head nerve stimulation, indicated the necessity of these cells for the expression of directed siphon responses to tail or head stimulation, respectively. 7. Because of their apparent necessity and sufficiency for directional siphon responses to anterior and posterior stimulation, these few cells provide well-defined vantage points for studying neural mechanisms underlying the motor control and transformation of siphon responses. The four LFSB cells offer a special advantage for cellular analysis because they form a homogeneous functional unit in which any sampled LFSB cell can be used as a precise monitor of the total motor output underlying the siphon T response.(ABSTRACT TRUNCATED AT 400 WORDS)

1. Little is known about modulation of action potential discharge in Aplysia mechanosensory neurons during defensive responses. The present studies examined rapid effects of noxious stimulation (occurring within 0.5-10 s) on the number of action potentials evoked by test stimuli delivered to the tail. Responses were monitored in the somata of mechanonociceptors in the pleural ganglion. A major hypothesis to be tested was that an important function of previously described alterations of membrane conductances in the sensory neuron soma is to generate an after-discharge that amplifies sensory signals during severe noxious stimulation of the cell's receptive field. 2. Discharge of spikes evoked by a moderate tap to one part of a sensory neuron's receptive field on the tail was enhanced by strong shock delivered 10 s earlier to another part of the field. Part of this enhancement appears to be due to a decrease in conduction block in central regions of the sensory neuron. 3. Repeated delivery of innocuous, moderately intense tail shock at 5-s intervals caused a progressive increase ("wind-up") of discharge, whereas repeated delivery of weak tail shock had no significant effect on discharge. In some cases the increase in action potential number involved a buildup of afterdischarge. 4. A single strong tail pinch sometimes induced an afterdischarge lasting < or = 2 s. Afterdischarge could also be induced in the isolated nervous system by intense electrical stimulation of the nerve containing the sensory neuron's main axon. 5. Several observations suggest that afterdischarge requires cooperative effects of a relatively large number of coactivated fibers in the test pathway. In contrast to pinching stimuli (which stimulated a larger part of the tail), intense, punctate stimulation with von Frey hairs failed to produce afterdischarge. Weaker tail or nerve stimulation failed to produce afterdischarge, even when short-latency, high-frequency discharge was evoked in the sensory neuron. 6. Cooperative effects on afterdischarge may differ from those involved in activity-dependent enhancement of presynaptic facilitation because simultaneous pairing of high-frequency activation of a single test sensory neuron with strong stimulation of a peripheral nerve lacking an axon of the tested sensory neuron was not sufficient to produce afterdischarge. The cooperative effects on afterdischarge may function to encode information about both the severity and spatial extensiveness of an injury. 7. Artificial hyperpolarization of the soma often reversibly reduced or abolished afterdischarge evoked by stimulating the nerve or tail. Thus the afterdischarge is often generated in or near the sensory neuron soma.(ABSTRACT TRUNCATED AT 400 WORDS)