Electrophysiology of CTA
Abstract and Keywords
Gustatory CS is analyzed by taste cells, the output fibers of which form chorda tympani (CT) and glossopharyngeal nerve (GN) to area postrema (AP). The labelled line (LL) and across nerve pattern (ANP) or across region pattern (ARP) theories assume that the unpleasant taste corresponds to the sum of discharging neurons or to the IC surface activated by c-fos marking. This information is transmitted by vagus nerve and by C fibers of the splanchnic nerve to area postrema. CTA changes the CS to a quinine-like bitter taste usually projected to external lateral subnucleus of PBN. This changes the hedonic values of the CS from attractive to aversive. The increase reflects the change of taste quality in the basolateral amygdala (BLA) and hedonic change in the central amygdala (CeA). Lesions of BLA disrupt CTA by interruption of BLA projections to the ventromedial hypothalamus (VMH).
Electrophysiological studies provide, together with behavioural effects of lesions and pharmacological experiments, essential information about the neural mechanisms of CTA. This chapter describes the processing of neural information in the projection sites of gustatory (CS) and visceral (US) senses. Although there is no direct electrophysiological evidence for the occurrence of associative learning between gustatory and visceral inputs in single neurons, altered responsiveness to the CS as a result of taste aversion learning has been reported at different levels of the gustatory pathway.
6.1 Taste qualities
It is generally agreed that there are four basic tastes, i.e. sweet, salty, sour, and bitter. In addition to these four tastes, recent studies allow us to include another taste, ‘umami’, into the basic taste category (Kawamura and Kare 1987 ; Kawamura et al. 1991 ). Umami is a Japanese word used by Ikeda ( 1909 ) to describe the taste of monosodium glutamate (MSG) which he found to be the essence of the savoury taste of sea tangle. An increasing number of taste researchers are now accepting umami as the fifth basic taste.
These basic tastes are also recognized by different species of mammals with a well known exception that cats do not have taste fibres responsive to sweeteners (Pfaffmann 1941 ). There are also species differences for the umami taste, i.e. hamsters are almost lacking receptors for umami substances such as MSG and inosine 5-monophosphate (IMP) (Yamamoto et al. 1988 ), rats cannot discriminate umami from sweet tastes (Yamamoto et al. 1991 ), but electrophysiological studies suggest that monkeys can discriminate umami from the other four basic tastes (Baylis and Rolls 1991 ). Behavioural support for the assertion that mammals can discriminate each of the basic tastes is derived from experiments using the CTA paradigm (e.g. Ninomiya et al. 1984 ; Frank and Nowlis 1989 ; Yamamoto et al. 1991 ) (Fig. 6.1 ).
The taste buds, the specialized end organs for gustation, are embedded in the stratified epithelium of the tongue, soft palate, pharynx, larynx, and epiglottis in mammals (Fig. 6.2 ). Lingual taste buds are associated exclusively with characteristic papillae such as the fungiform, foliate, and circumvallate papillae, while those outside the tongue are found on smooth epithelial surfaces. Each taste bud, which is about 70 μm high and has a diameter of 40 μm, consists of about 100 cells arranged in a compact, barrel-shaped (p. 77 )
The concept of the five basic tastes may be supported by five different transduction mechanisms corresponding to each of the five basic taste stimuli, e.g. sucrose, NaCl, HCl, quinine hydrochloride, and MSG for sweet, salty, sour, bitter, and umami, respectively. Each taste modality has its own specific mechanism (Kurihara et al. 1994 ): salty taste is mediated by Na+ influx through apical and basolateral Na+ channels; sour results from H+ blockade of K+ channels; sweet is mediated by activation of guanine nucleotide binding protein (G-protein ) which is coupled to receptors, leading to adenylyl cyclase-mediated elevation of cAMP; bitter is also transduced via a G-protein -dependent mechanism involving Ca2+ release from internal stores; umami may be mediated by metabotropic glutamate receptors.
In laboratory experiments, we use a taste CS with a simple taste without odours to establish CTAs. In the natural environment, however, animals as well as humans encounter various foods, most of which have complex tastes (p. 79 ) and odours. Therefore, they may use complex mixtures of tastes and/or odours of food components as sensory cues to acquire learned food aversions. This is the reason why they show food-specific aversions rather than taste-specific aversions. Such a conclusion is supported by Bernstein ( 1978 ) who reported that cancer patients receiving chemotherapy exhibited aversions to an unusual ice cream eaten just before the drug treatments, but that the aversion did not generalize to other unusual ice creams, which means that the patients did not use sweet: taste contained in the ice cream as the sensory cue in the aversion learning, but a complex taste possibly combined with odours.
6.2 Central gustatory pathway
The first step of initiation of taste sensation is stimulation of the taste receptor cells in the taste buds by ions or molecules of food substances and generation of receptor potentials which induce impulse discharges in the four taste nerves (Figs 6.2 and 6.3 ): the chorda tympani (CT), glossopharyngeal nerve (GN), greater superficial petrosal nerve (GSP), and superior laryngeal nerve (SL). The CT innervates the taste buds in the fungiform papillae on the anterior two-thirds of the tongue; the lingual branch of the GN, those in the foliate and circumvallate papillae on the posterior third of the tongue; the GSP, the palatal taste buds; and the SL, the taste buds on the pharynx, larynx, and epiglottis.
A schematic drawing of the central gustatory pathway in the rat is shown in Fig. 6.3 . The taste nerves terminate rostrocaudally in the NTS, i.e. the CT and GSP terminate most densely in the rostral pole, and the GN and SL project more posteriorly from the CT/GSP projection zone extending to the area postrema (Torvik 1956 ; Beckstead and Norgren 1979 ; Contreras et al. 1982 ; Hamilton and Norgren 1984 ). In rats, neurons in the gustatory (rostral) zone of the NTS project to the ipsilateral medial PBN, while those in the viscerosensory (caudal) zone project to the ipsilateral lateral PBN (Norgren 1978 ; Ricardo and Koh 1978 ; Herbert et al. 1990 ). The VPMpc of the thalamus, which receives taste inputs from the bilateral PBN, is the synaptic relay of the ascending pathway to the gustatory area of the IC (Yamamoto et al. 1980 ; Yamamoto 1984 ; Kosar et al. 1986 ; Cechetto and Saper 1987 ; Ogawa et al. 1990 ). On the other hand, axons from ipsilateral PBN pass ventrally beneath the thalamus giving off fibres to the lateral hypothalamus. They end in the central nucleus of the amygdala and in the bed nucleus of the stria terminalis (Norgren 1976 ). In monkeys, only the visceral caudal half of the NTS sends axons to the PBN, whereas neurons from the gustatory area of the NTS directly reach the VPMpc and then terminate in the IC (Kinney 1978 ; Beckstead et al. 1980 ). (p. 80 )
6.3 Processing of taste information
Peripheral taste nerve fibres or central taste-responsive neurons at different levels of the taste pathways do not respond specifically to only one of the basic tastes. Each unit usually responds to more than one of the basic tastes. Such multiple responsiveness to the basic tastes leads to the formulation of the across-neuron pattern (ANP) theory (Erickson 1963 ). According to this theory, taste quality is coded by the relative amount of activity generated across many neurons.
On the other hand, on the basis of the finding that individual fibres of the hamster and monkey chorda tympani may be categorized into three or four groups that show a best response to one of the four basic taste stimuli, Frank ( 1973 ) and Pfaffmann ( 1974 ) proposed the labelled-line (LL) theory. According to this theory, information about taste quality is carried by a subgroup of neurons that show greater responsiveness to one of the basic (p. 81 ) taste qualities than to the others. Recent findings indicate that responses of taste nerve fibres of the chimpanzee tend to respond exclusively to one of the five basic tastes, that gymnemic acid which inhibits sweet taste sensation in humans suppresses responses of only sucrose-best fibres (or fibres which show the most predominant responses to sucrose among the five basic tastes), and that amyloride which inhibits salty taste suppresses responses of only NaCl-best fibres (Hellekant and Ninomiya 1991 ). These results suggest: that each of the basic tastes is conveyed via specific lines of fibres.
Neural substrates for the LL theory may include a topographical arrangement of clusters of neurons with similar response characteristics to taste stimuli. An example of such chemotopical segregation of neurons is suggested in the IC of the rat, i.e., sucrose responses are dominant in the anterodorsal region, NaCl responses in the central and ventral regions, quinine responses in the posterior region, and HCl responses are distributed evenly in the gustatory region of the IC (Yamamoto et al. 1985a ). More recently, Yamamoto et al. ( 1994a ) have elucidated the functional segregation within the rat PBN by use of c-fos immunohistochemistry for anatomical marking of activated neurons: neurons in the external lateral and external medial subnuclei receive taste information for bitter and sour stimuli, neurons in the central lateral subnucleus are related to sweet taste, and neurons in the central medial subnucleus to salty taste (see Chapter 7 for details).
The two theories are not completely different, but each of them seems to stress one aspect of common taste-elicited neural responses, i.e. the importance of a subset of neurons in the processing of taste information (the LL theory) versus the importance of the entire population of neurons (the ANP theory). The above-mentioned effects of gymnemic acid and amyloride can also be explained by the ANP theory, because activation of a specific group of neurons is necessary for the formation of the acrossneuron pattern specific to a taste stimulus, Inactivation of a group of neurons also deteriorates an across-neuron pattern, which leads to diminution or alteration of the taste quality. A conceivable interpretation of the two quality coding principles suggests that a gross discrimination of taste categories is quickly achieved by the LL mechanism (Halpern and Tapper 1971 ) and the subtle difference of taste quality within the same modality is processed in a more time-consuming manner by the ANP mechanism (Scott 1974 ).
The existence of chemotopic organization in the cortical gustatory area of the rat prompted Yamamoto et al. ( 1985a , b ) to propose an across-region pattern (ARP) mechanism for taste quality coding at the cortical level, i.e. taste quality is coded by the pattern of the relative amount of neural responses across anatomically discrete areas rather than across individual neurons. The ARP concept explains the qualitative similarity among the taste stimuli found in the behavioural experiment using the CTA paradigm better (p. 82 ) than the ANP or LL notions (Yamamoto et al. 1985b ). The ARP hypothesis seems to include the key idea of the other two hypotheses. First, the assumption that the chemotopic organization of the cortex represents terminal projection sites of specific channels that convey each of the basic taste qualities leads to the conclusion that the ARP hypothesis accommodates the basic features of the LL hypothesis. Second, the ARP hypothesis may be considered a modified version of the ANP theory since the ARP notion assumes that taste quality is coded by the pattern of the relative amount of neural responses across anatomically discrete areas rather than across the individual neurons as the ANP theory asserts.
6.4 General visceral information
Lithium chloride (LiCl) is one of the most commonly used USs in the CTA paradigm. According to Nachman and Ashe ( 1973 ) and Nachman and Hartley ( 1975 ), an intraperitoneal (i.p.) injection of isotonic (0.15 M) LiCl (2% of body weight) following a gustatory CS is very effective in establishing strong CTA in the rat. Signs of sickness such as decreased activity, diarrhoea, and urination appear within 5–10 min after injection of LiCl (Nachman 1970 ; Nachman and Hartley 1975 ; Di Lorenzo 1985 ). An electrophysiological study (Yamamoto et al. 1989 ) showed that neurons in the IC of the rat responded tonically to an i.p. injection of 0.15 M LiCl with the shortest latency being approximately 5 min. These findings suggest that LiCl exerts its effect on the peripheral gastrointestinal organs and central nervous system with the latency of 5–10 min.
Niijima and Yamamoto ( 1994 ) have recently shown in anaesthetized rats that direct application of 0.15 M LiCl into the intestines elicits gradually increasing and long-lasting discharges in afferent fibres of the vagal (parasympathetic) and splanchnic (sympathetic) nerves with an onset latency of 5–10 min. Their results yielded the following findings. (1) The splanchnic and vagal nerves respond to LiCl administered not only intraperitoneally but also intraduodenally. This finding explains the fact that drinking of LiCl as well as an i.p. injection of LiCl exert US effects (Nachman 1963 ). (2) The magnitude of neural discharges is larger in the splanchnic nerve than in the vagal nerve both after intraduodenal and i.p. application of LiCl, suggesting a more important role of the splanchnic nerve than of the vagal nerve in transmitting the LiCl-induced sensory information. These results are in agreement with the finding by Martin et al. ( 1978 ) that bilateral subdiaphragmatic vagotomy in rats did not prevent subsequent acquisition of CTAs when toxicosis was induced with isotonic LiCl. (3) The LiCl activation of C-fibres rather than of B-fibres suggests the nociceptive nature of the information transmitted.
(p. 83 ) The activation of nociceptive fibres by LiCl application corresponds well with the recent c-fos immunohistochemical study showing that after i.p. injections of LiCl (Yamamoto et al. 1992a , b ) or acetic acid (Menetrey et al. 1989 ), c-fos immunoreactive neurons were found in the superficial dorsal horn, a projection site for visceronociceptive information in the thoracolumbar spinal cord (Sugiura et al. 1989 ). This spinal visceronociceptive information is known to project to the PBN (Hylden et al. 1989 ), a nucleus also receiving gustatory information.
Adachi and Kobayashi ( 1988 ) and Adachi et al. ( 1991 ) demonstrated that some AP neurons responded to direct application of isotonic LiCl. They suggest that excitation of AP neurons by LiCl through a humoral route plays a role in establishing CTA. Ritter et al. ( 1980 ) and Ladowsky and Ossenkopp ( 1986 ) showed that lesions of the AP impaired lithium-induced taste aversion.
Taken together, intraperitoneally injected LiCl exerts its effects as an US with a latency of 5–10 min through the neural routes via the vagus and splanchnic nerves and also through a humoral route.
Recent studies of the visceral sensory system indicate representation of multiple visceral modalities at all levels of the central nervous system (see Cechetto 1987 for review). In the NTS, general visceral afferents are represented caudally, and gustatory afferents rostrally (Adachi 1984 , 1994 ; Chambert et al. 1993 ). At the pontine relay, the PBN, general visceral afferents are represented laterally, but gustatory afferents medially (Cechetto 1987 ; Yuan and Barber 1991 ; Suemori et al. 1994 ).
Suemori et al. ( 1994 ) have shown in the rat that neurons in the lateral PBN respond to visceral stimulation such as distension of the stomach. Yuan and Barber ( 1991 ) recorded unit responses in the lateral PBN of the cat to electrical stimulation of gastric vagal and greater splanchnic nerves and suggested that this nucleus receives and processes a substantial amount of general visceral afferent input. The medial PBN (gustatory zone) projects to the parvocellular part of the ventral posterior medial thalamic nucleus, while the more lateral portions of the PBN (general visceral zone) project to the parvocellular part of the ventral posterior lateral nucleus of the thalamus (Cechetto and Saper 1987 ). These thalamic relays in turn project to the insular cortex. Single-unit recordings in the insular cortex of the rat have demonstrated that neurons in the insular cortex respond to multiple visceral sensory modalities and that there is a viscerotopic organization such that a rostral dysgranular insular zone receives primarily gustatory information whereas general visceral sensation is represented in the caudal granular insular region (Cechetto and Saper 1987 ; Ito 1992 ). The hypothalamus and amygdala have also been shown to receive direct viscerosensory projections from the NTS and PBN in the rat. The lateral PBN provides substantial innervation of the central nucleus of the amygdala (Saper and Loewy 1980 ).
(p. 84 ) As described above, gustatory information and abdominal visceral information ascend closely together so that it is possible for both inputs to converge on single neurons at all levels of the neuraxis. Recording sites of single unit responses to visceral stimulation seem to overlap with the gustatory responses in the thalamus (Saleh and Cechetto 1993 ) and in the IC (Ito 1992 ). Actual convergence of gustatory and abdominal visceral inputs has been reported in the amygdala (Yamamoto and Fujimoto 1991 ) and IC (Yamamoto et al. 1989 ). Hermann and Rogers ( 1985 ) recorded unit responses of the rat PBN to electrical stimulation of the vagal and taste nerves and found convergence of both inputs onto neurons within an interstitial zone in the caudal PBN. The above electrophysiological studies showing projections of gustatory and general visceral signals to the same common levels of the neuraxis together with the recent c-fos immunohistochemical studies and behavioural lesion studies have considerably advanced our knowledge of the neural substrate of taste-visceral integration underlying the formation of CTA.
6.5 Alteration of taste responsiveness after CTA
On the basis of electrophysiological characteristics of taste-responsive neurons, it is of great interest to survey changes of taste responsiveness of central neurons to the CS after acquisition of CTA. A specific question is whether CTA learning is accompanied by a change of taste quality of the CS or only by a change of hedonic appreciation of the CS (i.e. the taste quality of the CS is not altered but the CS elicits aversive behavioural responses after CTA).
Since the pioneering studies done by Aleksanyan et al. ( 1976 ) and Buresova et al. ( 1979 ), changes of neural responses after the acquisition of CTA have been reported at different levels of neuraxis of the gustatory pathway such as the NTS (Chang and Scott 1984 ), PBN (Di Lorenzo 1985 ; Shimura et al. 1997b ), IC (Yamamoto et al. 1989 ), and amygdala (Yamamoto and Fujimoto 1991 ; Yasoshima et al. 1995a ). The neural mechanisms mediating the response modifications in CTA-trained rats will be a key to understanding the neural basis of CTA. Many of these studies suggest an enhanced response to the CS after the acquisition of CTA, suggesting that the increased responsiveness to the CS is useful for detecting low concentrations of the CS to avoid its ingestion. However, as already suggested by Buresova et al. ( 1979 ), inhibition of unit activity depending on the location and function of neurons is also frequently induced by presentation of the CS after aversive conditioning.
(p. 85 ) 6.5.1 NTS
Chang and Scott ( 1984 ) recorded unit responses of the NTS to taste stimuli in CTA-trained rats and control rats under anaesthetized, paralysed conditions. They found that the saccharin CS evoked a significantly larger response in CTA-trained rats than in control rats, and that the effect was limited to the subgroup of neurons sensitive to sweeteners (Fig. 6.4 ). These results imply the existence of a functionally distinct channel within the taste system, which is composed of neurons responsive to the sweet quality. Moreover, these alterations of response patterns of the NTS neurons
Does the taste quality of saccharin change to a bitter or quinine-like taste after aversion learning to saccharin? This question is raised since saccharin contains a bitter component (Bartoshuk 1979 ) and c-fos immunoreactive labelling induced by saccharin shifts from the dorsal lateral subnucleus, where sweeteners elicit c-fos expression, to the external lateral subnucleus, where quinine elicits c-fos expression, as the result of learned aversion to saccharin. Chang and Scott ( 1984 ) showed that within the three-dimensional taste space, as determined by neural activity in the NTS, the saccharin CS after the acquisition of CTA was in close alignment with quinine. It should be (p. 86 ) noted here that the taste spaces represent similarities among stimuli in terms of taste hedonics as well as of taste quality. The same paper reported that the CS evoked a significantly larger response in CTA-trained rats than in control rats, and that the effect was limited to the subgroup of neurons sensitive to sweeteners. On the basis of these results, Scott and Giza ( 1987 ) have suggested that among the gustatory processes performed in NTS it is primarily the hedonic appreciation of a taste, rather than its quality, that appears to be altered. In fact, behavioural studies have shown that aversion learning to saccharin generalizes only to sweeteners, and not to other taste stimuli representing salty, sour, and bitter tastes, indicating that the taste quality of saccharin does not change from ‘sweet’ to ‘bitter’ (Nowlis et al. 1980 ; Frank and Nowlis 1989 ).
It is suggested that a subset of PBN neurons shows enhanced responses to a taste stimulus which has been paired with LiCl administration. Di Lorenzo ( 1985 ) examined the effects of taste aversion conditioning on unit responses to NaCl in the PBN of rats paralysed with Flaxedil. Initially, units were tested for responsiveness to NaCl, HCl, sucrose, saccharin, and quinine, then units were conditioned by a presentation of NaCl overlapping with an i.v. injection of LiCl. This conditioning procedure resulted in a long-lasting overall increase in the magnitude of the response specifically to NaCl in a subset of PBN units. The conditioned units were more narrowly tuned than units that were not conditioned, but did not necessarily respond best to NaCl. The temporal pattern of response to NaCl did not change after conditioning, i.e. both phasic and tonic portions of the NaCl response increased in magnitude.
Shimura et al. ( 1997b ) examined the nature of parabrachial responses to 13 taste stimuli in deeply urethane-anaesthetized rats that were pretrained to avoid 0.1 M NaCl in a behavioural paradigm. These responses (n= 53) were compared to the responses (n = 56) of naive rats. In both groups more than 80% of units were sensitive primarily to NaCl. Although there were no differences between the mean magnitudes of the responses to each taste stimulus between groups, cluster analysis clearly revealed a strong similarity among responses to sodium salts in the conditioned group (Fig. 6.5 ). These results suggest that the aversive conditioning modifies PBN units so that salty taste is more salient than other tastes. This would facilitate the gustatory discrimination that is required by conditioned animals.
The existence of centrifugal control of PBN neuronal activity is suggested by the findings that decerebration (Di Lorenzo 1988b ), infusions of procaine into the IC (Di Lorenzo 1990 ), or electrical stimulation of the IC (Kiyomitsu et al. 1988 ) all influence responsiveness of PBN neurons. Di Lorenzo's (p. 87 )
(p. 88 ) 6.5.3 Insular cortex
Based on the responsiveness of neurons in the IC to various taste stimuli in behaving rats, taste-responsive neurons are grouped into a taste-quality type, which reflects the quality coding, and a taste-hedonic type, which differentiates the hedonic (preferable or repulsive) aspect of taste stimuli (Yamamoto et al. 1989 ). Responses of some of the cortical taste-responsive neurons are modified after acquisition of CTA produced by pairing a taste solution with an intraperitoneal injection of LiCl (Yamamoto et al. 1989 ). The effects are different, depending on the type of neuron. The enhanced responses were observed in taste-quality neurons, which are narrowly tuned neurons responsive specifically to a group of tastants eliciting one of the conventional four taste qualities, and a preference response pattern changed to an aversive response pattern after acquisition of CTAs in taste-hedonic neurons, which are defined as neurons showing differential response patterns depending upon hedonics (preference or aversion, acceptance or rejection) of taste stimuli.
More recently, Yasoshima et al. ( 1995b ) studied the role of the IC in CTA of freely behaving rats by comparing the pre- and post-conditioning unit activities to intraoral infusions of saccharin which was used as a CS. After recording the control responses, rats drank CS for 5 min followed by an i.p. injection of LiCl as a US. At 10 min after the US injection, the rats did not show aversive behaviour to the CS and the CS elicited no significant unit activities. However, 30 min after conditioning, rats showed aversive taste reactivities to the CS and significantly enhanced activities to the CS were observed in 14 (26%) out of 54 units. The enhanced responses in 6 of the 14 units were maintained for more than 6 h. On the other hand, in the other 8 units, the enhanced responses were present for no more than 90 min. These two different time courses of excitatory changes suggest that the IC neurons are involved in formation of both short-term and long-term CTA memories.
The amygdala plays a crucial role in the judgement of hedonic values of sensory stimuli in some instances of emotional learning (LeDoux 1993 ; Muramoto et al. 1993 ; Helmstetter and Bellgowan 1994 ). Because the hedonic value of the CS is drastically changed from ingestive to aversive after the acquisition of CTA, the amygdala is assumed to be one of the most plausible sites involved in CTA formation. To investigate the dynamic neural processing within the amygdala in the context of the CTA paradigm, Yamamoto et al. ( 1993a ) and Yasoshima et al. ( 1995a ) recorded amygdalar unit activities during voluntary licking or during intraoral infusions of taste solutions, and compared them before and after the conditioning procedure in freely behaving rats (Fig. 6.6 ).
According to their results, only a limited number of amygdalar units (9 of 137 units)were responsive to taste stimuli before conditioning. The US (an i.p. injection of 0.15 M LiCl) elicited excitatory and/or inhibitory responses in another 9 of 137 units. These responses lasted less than 2 h. These LiCl-responsive units did not respond to any taste stimuli before and after conditioning. When tested 3–4 h after pairing saccharin (CS) with the US, 32 of 137 units responded to the CS: 20 of these 32 units were facilitatory-type and 12 were inhibitory-type (Fig. 6.7 ).
The facilitatory-type units, which responded to the CS in an excitatory direction after CTA, were found more frequently in the BLA (n=17) than in the CeA (n=3). The facilitatory-type units showed a response pattern similar to that of the taste-quality type. It is suggested that the BLA plays important roles in aversive or avoidance behaviour, such as neophobia (Nachman and Ashe 1974 ), fear conditioning (LeDoux 1993 ; Helmstetter and Bellgowan 1994 ), passive (Yajeya et al. 1991 ) and discriminative (Maren et al. 1991 ; Muramoto et al. 1993 ) avoidance learning, and CTA (Yamamoto et al. 1995a ). The facilitatoy responses in the BLA, therefore, may be related to the conditioned aversive responses, but not to such innate aversive reflexes as elicited by unpalatable stimuli, HCl or QHCl, because the latter stimuli did not induce any remarkable responses in these units.
On the other hand, the inhibitory-type units, which responded to the CS in an inhibitory direction after CTA, were found more frequently in the CeA (n=11) than in the BLA (n=1). Aversive taste stimuli such as HCl and QHCl also inhibited the spontaneous activity of these units regardless of the CTA procedure, indicating that these units belong to the taste-hedonic type. The occurrence of such inhibitory responses to the hedonically negative tastes suggests that the normal function of the CeA is to stimulate ingestive behaviour. This suggestion is supported by the recent finding (Minano et al. 1992 ) that inhibition of CeA neural activities by microinjection of muscimol, a GABA-selective receptor agonist, into the CeA produced a decrease of food intake. Alternatively, these inhibitory units might be related to the taste-elicited autonomic responses because it is known that the CeA is involved in the autonomic conditioned responses such as heart rate conditioning (Kapp et al. 1992 ) and conditioned cephalic insulin response (Roozendaal et al. 1990 ).
The above-mentioned electrophysiological findings suggest that recognition of the CS and expression of rejection behaviour after the acquisition of CTA is based on an interplay between the BLA and CeA, i.e. qualitative and cognitive aspects of the CS are processed in the BLA and hedonic aspects in the CeA. Behavioural lesion studies have shown that lesions of the BLA (p. 91 ) severely disrupt the acquisition and retention of CTA, while lesions of the CeA have no effect on the CTA formation (Yamamoto et al. 1995a ). Since the BLA has dense excitatory fibre connections to the ventromedial nucleus of the hypothalamus (VMH) (Oomura and Ono 1982 ), which is known to be the ‘satiety centre’, excitation of VMH from the BLA may play a role in avoidance behaviour. A more important role of the BLA is to evaluate the biological significance of sensory stimuli for the animals (Ono and Nishijo 1992 ). Without the BLA, therefore, CTAs are disrupted because the animals cannot associate the CS with the US during acquisition or recognize the aversive nature of the CS during retrieval.
Hypothalamic neurons also change their response patterns in terms of the hedonic dimension, i.e. saccharin (CS) activates the feeding centre and inhibits the satiety centre in naive animals, but elicits opposite effects after CTA acquisition to the CS (Aleksanyan et al. 1976 ). Since the hypothalamus receives inputs from the IC and the amygdala, altered responsiveness of hypothalamic neurons may simply reflect that of IC and/or amygdala neurons which as described above exhibit altered responses to the CS after CTA acquisition.