Section 3

Learned Food Aversions

3.0 Introduction

The key message of Section 2 is that food rejection behaviour is multidimensional. By comparing the different facets to the rejection of different items, we can construct a useful classificatory scheme. However, those different facets to the rejection behaviour may be of interest in and of themselves. In this Section, I look at one such case in detail, that involving learned food aversions.

Learned food aversions (LFAs) are well known from the animal literature: a food followed by any of various different conditions, notably gastro-intestinal malaise, is subsequently avoided. They are a very important model for our understanding of food-related learning behaviours and for learning behaviours in general. LFAs have also been referred to as "conditioned taste aversions", as well as "conditioned flavour aversions", "learnt taste aversions", "conditioned food aversions", "bait shyness" and so on. I avoid using these terms as the aversion is not necessarily produced through a conditioning process and as taste is not the only sensory modality that can carry such an aversion.

This Section will investigate the precise nature of the response induced by an LFA, specifically whether this involves the target item becoming nauseating. That is, does a learned food aversion involve a learned nausea?

3.1 Literature Review

LFAs, it now appears, are involved in many more phenomena than the early psychological researchers in the field might have realised. This multiplicity of sources of evidence concerning LFAs can be of great help in considering theories about LFAs, although the uncertainty about which phenomena are related—and quite how they are related—to LFAs may also confuse.

The exact requirements needed to produce LFAs remain unclear. LFAs require changes in the milieu interne, but not necessarily nausea, despite research concentrating on nausea. There seems to be a unifying factor in the various possible external stimuli involved in LFAs in their relationship to food. For example, olfactory cues are potentiated as stimuli in LFA learning in rats when they are paired with tastes, i.e. when they were part of an eating experience (Rusiniak, Palmerino, Rice, Forthman & Garcia, 1982; Garcia, 1989). Wilcoxon, Dragoin & Kral (1971) found visual cues served as stimuli in bobwhite quails (Colinus virginianus); a species, as they point out, that primarily uses vision in its food selection behaviour, compared to the rat, using odour and taste. In humans, taste seems to be the major relevant sensory modality though smell, texture, appearance and even sound may also carry aversions (Logue, Ophir & Strauss, 1981).

There has been a tremendous amount of research on some aspects of LFAs. Experiments involving pairing nausea with a flavoured drink in rats are legion and the production of a LFA is commonly used as an experimental tool, as an index of nausea. Yet other aspects of LFAs have been less scrutinised. Studies in humans are rare. There is a paucity of experimental manipulations in humans, but a number have now demonstrated LFAs in controlled situations (Lamon, Wilson & Leaf, 1977; Cannon, Best, Batson & Feldman, 1983; Arwas, Rolnick & Lubow, 1989; Okifuji & Friedman, 1992, Hu, Willoughby, Lagomarsino & Jaeger, 1996).

This section concentrates on investigating the nature of the aversion produced in a LFA. Following from Section 2, I seek to understand how the target of a LFA fits into the taxonomies described. In particular, I will ask whether the target food becomes nauseating, something less considered in the use of LFAs as paradigmatic of palatability change.

LFAs were first noted by zoologists in the eighteenth century. It was realised that warning markings adopted by some insects served to protect them, because predators would have previously associated those markings with either an unpalatable taste (what psychologists might describe as evaluative conditioning) or with the induction of GI malaise (an LFA). Other food rejections, as with certain plants, were also understood under the same paradigm. Psychology discovered LFAs for itself a number of times, beginning with a series of studies by Garcia and colleagues using the favoured experimental animal, the rat.

3.1.1 Experimental Psychology

3.1.1.1 Non-humans

Garcia and colleagues were investigating learning phenomena in rats when they discovered LFAs (see Garcia, 1989 for review). Expecting to find classical conditioning, they found that pairing X-rays (which produce nausea) with a food showed a novel pattern of learning, the LFA. Research has continued to concentrate on the rat since then. The prototypical paradigm for a LFA can be described as:

A taste (or other food-related stimulus) followed by gastro-intestinal (GI) malaise, typically induced through drugs or X-irradiation, leads to an aversion for that taste. LFAs were used to argue against an equipotentiality in learning and for a concept of preparedness. A cynical observer might describe this as an attempt to shoehorn LFAs into a classical conditioning model, as opposed to simply viewing them as a distinct learning phenomenon.

The significance of LFAs was expanded when Rozin (1968; 1976) argued that the anorexias seen in rats fed specific nutrient-deficient diets were actually due to LFAs. Rozin (1968) demonstrated that the rats actually developed an aversion specific to the nutrient-deficient diet on which they were fed as opposed to a general anorexia. Many nutrient deficiencies result in chronic nausea, thus the diet would become repeatedly paired with nausea, producing a LFA. With diets containing many items, all components of the diet become repeatedly paired with nausea, producing a suite of LFAs.

A similar situation was shown to arise in "tumour anorexias" in rats (Bernstein & Sigmundi, 1980; Bernstein & Fenner, 1983; Bernstein, Treneer, Goehler & Murowchick, 1985). The tumour produces a chronic state that results in LFAs developing to whatever diet the rat is on. The same may be occurring in the anorexias following a number of surgical procedures, like vagotomies (Bernstein & Goehler, 1983) and intestinal bypasses (Bernstein & Borson, 1986).

There has been considerably less research done on other species, but the same paradigm has been shown to produce LFAs in numerous other species, throughout the mammals and birds and beyond, e.g. land slugs (Limax maximus; Gustavson, 1977) and mantises (Berenbaum & Miliczy, 1983). Whether such phylogenetically diverse occurrences (three different phyla) of the phenomenon are from a common ancestor or the product of convergent evolution (as Garcia, Lasiter, Bermudez-Rattoni & Deems, 1985, suggest) is another matter.

Early investigators had noted a division between stimuli sensed by extero- and interoceptors and suggested that it was specifically GI malaise that would produce LFAs (e.g. Garcia, Hankins & Rusiniak, 1974). The efficacy of X-irradiation, motion sickness and many drugs in producing LFAs can then be explained through their common property of causing GI malaise. However, a large number of other drugs that do not induce GI malaise have also been shown to produce LFAs. In fact, only a few drugs seem incapable of producing LFAs. GI malaise is a sufficient but not necessary factor. Gamzu (1977) has suggested that LFAs occur in response to any novel, internal state. Hunt & Amit (1987) considered the differences between LFAs in response to GI malaise-inducing drugs and procedures and those to drugs which, under other circumstances, are self-administered (SA) and do not appear to produce any GI malaise. They suggest that the two types of LFAs are qualitatively different and involve different psychological mechanisms. Heeding this suggestion, this thesis concentrates on LFAs induced by GI malaise and avoids the more contentious SA category of Hunt & Amit (1987).

Such non-GI malaise aversions are not reported in questionnaire studies in humans (e.g. Pelchat & Rozin, 1982 and see 3.1.1.2). Presumably this is because the necessary aversive stimuli (novel drug/internal states) are not normally encountered by most people, except perhaps recreational drug users (McAuliffle, 1975) or individuals suffering from certain diseases, notably cancers (see 3.1.2.3).

3.1.1.2 Humans

A number of studies (Lamon et al., 1977; Cannon et al., 1983; Arwas et al., 1989; Okifuji & Friedman, 1992) have used the same basic paradigm as in 3.1.1.1 to demonstrate the same psychological process of an LFA. That is, in experimentally controlled situations, these studies have produced an aversion to a food by pairing it with nausea.

Lamon et al. (1977) paired a target beverage with nausea by having subjects sip the target beverage while in an experimental set-up designed to induce motion sickness with the pseudo-Coriolis effect (see 3.4.1). However, errors elsewhere in the methodology give results which are difficult to interpret.

Cannon et al. (1983) induced nausea with injected apomorphine (controls receiving isotonic saline). Subjects drank 120ml of cranberry juice, as a novel food, before receiving their injection. They were then tested four days and one month after and an aversion was seen with some of the measurements taken. These consisted of consumption, ratings on 14 descriptive adjectives (alarming, unpleasant, dangerous, pleasant, relaxing, repulsive, distasteful, appetizing, safe, bad, tasty, harmless, good and harmful) and, at the first test session only, heart rate changes in response to the target (described in Baker, Cannon, Tiffany & Gino, 1984).

Arwas et al. (1989) used a different method of inducing motion sickness, prior to which subjects had been presented with a target beverage, either familiar (Coca-Cola or Sprite) or novel (Schweppes tonic or ginger ale). Consumption before and two hours after was measured. Once subjects who had reported no symptoms on the rotation procedure had been excluded, for an unfamiliar beverage (but not a familiar one), consumption decreased in the experimental group and not in a control group.

Okifuji & Friedman (1992) again used the pseudo-Coriolis effect. The target food ("nutmeg-ginger" ice cream) was presented immediately before this procedure, with the amount consumed and taste ratings being measured. At least 30 minutes afterwards, the amount consumed and taste ratings were again measured. Subjects unfamiliar with the target food rated it less palatable than controls and consumed marginally less of it.

In Hu et al. (1996), subjects were divided by whether they had tasted the target beverage, soybean milk, prior to the experiment. A procedure using the pseudo-Coriolis effect was used. Two hours subsequent, subjects again drank the soybean milk, with the amount consumed and a likableness rating recorded. Those who had not previously tasted soybean milk consumed less of it and rated it lower than those who had.

A number of retrospective questionnaire studies (Rozin & Fallon, 1980; Logue et al., 1981; Pelchat & Rozin, 1982; de Silva & Rachman, 1987; Pelchat, 1993) and anecdotal reports (most famously Martin Seligman’s Sauce Bearnaise phenomenon; Seligman & Hager, 1972) have reported the in vivo occurrence of LFAs in everyday life. These appear to fit the same LFA paradigm: a food followed by nausea leading to an aversion. However, there are obvious problems when away from the setting of a controlled experiment. There may be demand effects in the reporting of such occurrences. Individuals may feel the need, subconsciously, to justify their food dislikes by inventing explanations for their dislikes involving negative, post-ingestive effects, GI or other. Consider how the phrase "I can’t stomach it" confounds dislike with GI events.

3.1.2 Clinical (Human) Phenomena

3.1.2.1 Emetic Aversion Therapy

Early attempts at aversion therapy had assumed the law of equipotentiality, derived from the early work on classical conditioning. A preparedness concept was, however, independently discovered and emetic therapy was instituted for treating alcoholism (see Elkins, 1991, for review). This broadly follows the paradigm of 3.1.1.1, but with a number of important differences. The target "food" item, an alcoholic beverage, is very familiar and thus highly resistant to LFA formation, thus repeated pairings of the "food" and emesis are necessary; the subjects are not a representative sample (being alcoholics who wish to give up) and the alcohol itself may effect the psychological and physiological processes involved.

Disulfiram (better known under its trade name of Antabuse) is used in cases of alcoholism as it induces extreme malaise should the user concurrently consume any alcohol. As GI symptoms are included in its effects, the question arises whether disulfiram is effective through inducing a LFA to the taste of alcoholic beverages, as well as through more obvious manners. The efficacy of disulfiram has been questioned, but its failure rate seems to be down to low levels of compliance and, if its use is supervised, it does seem to aid in recovery (Brewer, 1992; Kristenson, 1992). While on Antabuse, patients do avoid drinking: only 24% would risk doing so in Brewer (1986). However, it is quite common for patients to alternate periods of relatively controlled drinking with periods on Antabuse (Brewer, 1992), arguing against any long-lasting aversion. Perhaps an LFA does develop but is readily extinguished.

Emetic aversion therapy has also been tried as a treatment for smoking and obesity. Again, as with alcoholism, the high familiarity of the target item requires repeated pairings. In tackling obesity, producing aversions to specific foods proves an unprofitable strategy. An aversion to, say, doughnuts has no effect on the rest of the diet and the individual may go on eating other high fat/calorie items.

Generally, emetic aversion therapy has fallen out of favour. The potential hazards of repeatedly inducing emesis and the obviously unpleasant nature of the treatment far outweighed its limited successes. In retrospect, it is worth noting that the emphasis on emesis was unnecessary and potentially misleading. It is nausea that produces LFAs. Obviously most methods of inducing emesis also produce considerable nausea, but clinicians could have perhaps avoided the specific problems associated with emesis.

3.1.2.2 Anticipatory Nausea and Vomiting (ANV)

Patients undergoing radio- or chemotherapy for cancer usually experience nausea and emesis as side-effects of their treatment. It is also common to experience these symptoms before an individual treatment session, thus anticipatory nausea and vomiting or ANV. This distressing phenomena need not have anything to do with food and has been viewed strictly in terms of "classical conditioning" (e.g. Challis & Stam, 1992). Triggers, conditioned stimuli, have been reported as including the hospital rooms or staff members. Conditioned emesis to non-food cues has also been demonstrated in animal models as long ago as Riddle & Burns’ (1931) work with pigeons. However, food triggers are often involved, resulting in a paradigm matching that in 3.1.1.1, though with repeated trials, an unrepresentative sample population and the possibility that the cancer may affect the processes involved.

ANV is not seen in all cases of chemotherapy (Bovbjerg, Redd, Jacobsen, Manne, Taylor, Surbone, Crown, Norton, Gilewski, Hudis, Reichman, Kaufman, Currie & Hakes, 1992), which would argue against an LFA interpretation given that LFAs develop with such ease. On the other hand, the involvement of nausea is suggestive. This is a thesis on food rejection, yet ANV demonstrates that there are important learning phenomena involving nausea which do not always involve food.

3.1.2.3 "Cancer Anorexia"

Many cancer patients show anorexia. This can clearly has multiple causes: e.g. concurrent nausea (either produced by the cancer or by treatment), depression, problems with swallowing (the salivary glands’ are inactivated by some radiotherapies; pain when swallowing is caused by many upper GI tract cancers) etc. As well as all these, it has been suggested that some "cancer anorexia" may parallel the tumour anorexia seen in rats, with the cancer or treatments thereof causing chronic or repeated nausea, leading to a suite of LFAs developing to the entire diet. The same may be occurring in the anorexias following some surgical procedures in humans, like vagotomies and intestinal bypasses, and even depression, all of which have chronic nausea as a symptom (see Bernstein & Borson, 1986, for review). In fact, anorexia is an extremely common symptom in hospital in-patients and probably an important cause of the high rates of malnutrition reported in hospitals (McWhirter & Pennington, 1994).

3.1.2.4 Pregnancy

Pregnancy is associated with both repeated feelings of nausea and with peculiar food cravings and dislikes. There is an obvious suggestion that these two are linked, but research into how much LFAs underlie pregnancy-associated dietary change is still inconclusive (Jarvis, 1990). Various other mechanisms have also been suggested to explain dietary changes from teleological ones concerning changing nutritional needs in pregnancy to physiological ones disturbing taste perception (Bowen, 1992; Jarvis, 1990). Conditioned nausea experiences to smells, rooms or other stimuli, paralleling ANV, are also anecdotally reported.

3.2 The Nature of the Response Produced

One of the major factors in food selection behaviour is palatability (Shepherd, 1989). The psychological mechanisms that determine a food’s palatability, however, are still largely unknown (Rozin & Vollmecke, 1986). One set of mechanisms that Rozin & Vollmecke discuss "involves special built-in abilities to alter preferences or likes in response to the delayed consequences of ingestion of a particular food" (p. 436). That the post-ingestional consequences of a food should affect subsequent behaviour towards that food is clearly of survival value, but whether a food’s post-ingestional consequences "alter... likes" for that food or, more specifically, its palatability is potentially a more contentious matter. There are clear examples where a food’s post-ingestional consequences affect behaviour towards it without affecting its palatability. One example in humans is from Pelchat & Rozin (1982) who, in a retrospective questionnaire study, found that negative skin reactions following ingestion of a food resulted in that food being rejected, but did not affect its palatability. There are fewer examples in animals, where it is far harder to study the nature of the aversion.

The central thesis behind Section 2 is that a food aversion can take several forms, not all necessarily associated with unpalatability. Where do learned food aversions (LFAs) fit? Rozin & Fallon (1980) included reference to an item where participants had to recall the target item of an accidental LFA in their life. This item fitted the proposed taxonomy of food rejections poorly, immediately suggesting that it is an important test of our theories in this regard. Specifically, the item was closest to their ‘distaste’ category, where the authors concluded it belonged, but was notable for also evincing nausea.

A ‘distaste’ categorisation follows the orthodox view that LFAs result in a hedonic shift in palatability. The target food is thought to become less pleasant in its taste. In other words, LFAs are the classical example of a mechanism for Rozin & Vollmecke (1986) that "involves special built-in abilities to alter preferences or likes in response to the delayed consequences of ingestion of a particular food". This gives us another reason why determining the nature of the aversion produced by LFAs is important. Moreover, LFAs are used as our main model of palatability change in general. If prior association with nausea results in a direct and simple drop in palatability, we have a psychological mechanism that directly affects palatability and which, therefore, may tell us more about food selection in general. Further still, LFAs form an important paradigm for our understanding of wider learning processes. This all means that the details of LFA learning are important, particularly whether LFAs alter the palatability of the target food or whether they produce avoidance of the target food in some other way.

If a hedonic shift in palatability does not solely produce the aversion seen in LFAs, what does? I will suggest that a learned nausea response develops to the target food; that is, that the target food becomes nauseating in itself. This brings us to a third reason why the nature of the aversion in LFAs is so important. This alternate hypothesis, as I will explain in 3.2.2, has very important implications for the role of LFAs in various clinical phenomena introduced earlier (3.1.2), like anticipatory nausea and vomiting (ANV).

Many people report LFAs and I have been collecting anecdotes during the course of my research, asking people how they feel on re-encountering the food in question. While of limited scientific value, their responses serve to introduce and illustrate some of the issues to be discussed later and provide some motivation for how we proceed. Below (3.2.1), I will present some more robust evidence.

What is noticeable about people’s anecdotal descriptions is that they refer to the target item as being nauseating. "And now the smell of Brussels sprouts cooking makes me retch," is a typical statement. Such statements were sometimes forthcoming before I asked any details, but when I did specifically ask about the nature of the aversion, there were even more responses matching this pattern and never any that would fit the orthodox hedonic shift idea.

Three anecdotal examples illustrate the range of related phenomena and show how LFAs may be related to other learned nausea phenomena. A general practitioner who had suffered nausea on taking a particular antibiotic now reported nausea when she encountered the drug. This fits the standard pattern for a LFA but reminds us that the target ‘food’ can be most anything, which can lead to quite unexpected consequences. This individual even reported feeling nausea when writing out a prescription for the drug, an extreme example of the potentiation of visual cues by association with ingestion?

A friend has an aversion to sticky price labels: they make her feel nauseous. She reports an incident in her childhood when she had been sick and remembers seeing a price label on a box of food. What at first seems a particularly odd phenomenon again appears to be a LFA to a particular cue that was potentiated by association with food.

A patient of a colleague complained of having vomited after eating clams. He described a first incident when he had been ill and vomited after eating clams, a food he had previously long eaten and enjoyed. Subsequently, when he had eaten clams, he had again vomited, but nevertheless reported that he continued to like clams very much. His doctor was unable to find any somatic cause of his reaction. Could this be an example of a learned nausea/emesis response to the food with complete disassociation from any hedonic shift in palatability?

3.2.1 Literature Review

The orthodox view in experimental psychology is that LFAs produce a hedonic shift in palatability. Numerous examples exist (e.g. Garcia, 1989; Rozin & Vollmecke, 1986; Baker, Cannon, Tiffany & Gino, 1984; Pelchat, Grill, Rozin & Jacobs, 1983; Pelchat & Rozin, 1982; Gustavson & Gustavson, 1982; Garcia et al., 1974; Kalat & Rozin, 1972) which have interpreted the consequences of LFA formation as below:

Nausea following a food leads to a drop in the palatability of that food, such that the reaction to the target food is then like that towards an innately unpalatable food. Compare, for example, Garcia & Hankins (1977), p.14: "Flavor-illness conditioning represents a motivational process in which the hedonic tone of a peripheral stimulus (flavor) is appropriately modified according to its usefulness in maintaining an internal homeostatic balance."

I suggest an alternate, but not necessarily mutually exclusive, explanation of the nature of the aversion:

Nausea following a food results in that food subsequently provoking a learned response of nausea, thus making the food aversive. Compare Zahorik (1972), p.190: "In this interpretation, the flavor which has been paired with illness actually elicits some of the symptoms of that illness, and changes in ingestion are only one of those symptoms."

Zahorik’s (1972) description casts the learned nausea in the role of a conditioned response. To use the terminology of classical conditioning, under this model, LFAs seen to be produced by pairing a taste—the conditioned stimulus (CS)—with nausea—the unconditioned response (UCR)—and then the subsequent avoidance of the same taste is due to its producing a conditioned response (CR), a copy of the UCR, i.e. nausea. The unconditioned stimulus (UCS) in this analysis would be whatever was originally used to produce the nausea. However, the relationship between classical conditioning and LFAs is very unclear (Seligman & Hager, 1972; Logue, 1979; Garcia, 1989), so it is unclear whether our knowledge of classical conditioning provides a useful model. What components of the LFA story are analogous to what components of classical conditioning is also debatable (Garcia, 1989). I avoid this issue and refer to a "learned nausea response", rather than necessarily a conditioned one. From the point of view of an organism’s reaction to the target of a LFA, it is irrelevant whether the nausea involved is conditioned or merely learned. The roles of nausea as "UCR" and "CR" may be wholly unrelated.

Hunt & Amit (1987) drew a distinction between LFAs caused by drugs that induce GI malaise and those which do not and are self-administered (SA) as described in 3.1.1.1. Learned aversions to drugs that do not produce nausea can clearly not produce a "conditioned" response of nausea. However, as research on SA-drug associated aversions is lacking, I am unable to comment much about the possible involvement of a learned nausea in the response to a target item.

Clearly, it is possible for both of these explanations, a hedonic shift and a learned nausea, to be true and I shall concentrate on trying to prove a learned nausea is involved rather than on proving or disproving the hedonic shift idea. Whether or not a hedonic shift also occurs, proof of a learned nausea response will require us to interpret the hedonic shift carefully.

When the first studies had made it apparent that LFAs were dissimilar in some ways from classical conditioning, the precise reaction of experimental rats to tastes previously paired with nausea was investigated. The reaction observed was compared to the reaction seen to innately unpalatable substances (prototypically quinine) and contrasted to the reaction seen to tastes previously paired with external injury (for review, see Rozin, 1976). For example, rats scatter food that has previously been paired with illness (Garcia et al., 1974) and Berridge, Grill & Norgren (1981) found mouth gaping, chin rubs, increased locomotion and fluid ejection in rats in response to sucrose solution previously paired with lithium. The reactions seen in both these studies are also seen to innately unpalatable food.

With reference to Hunt & Amit (1987) again, Parker (1982) later demonstrated different behavioural responses in rats to a food previously paired with lithium, which induces GI malaise, and to amphetamine, an SA drug that does not induce nausea. The amphetamine failed to elicit "chin-rubbing", one of the behaviours found to innately unpalatable foods.

Similarly, the response of rats to a thiamine-deficient diet—thought to also produce an LFA (see 3.1.1.1)—parallels their response to quinine, in both cases spilling of food (Rozin, 1967). Pelchat et al. (1983) found the orofacial responses of rats to a food previously paired with upper GI malaise were different to those to a food previously paired with shock. (Pairing a taste with lower GI malaise, induced by high levels of lactose, produced responses similar to those to foods previously paired with shocks.) All of which demonstrates that a food previously paired with internal discomfort is treated differently from one previously paired with external discomfort and, in some ways, like an innately unpalatable food. However, that the responses to a food previously paired with internal discomfort and to an innately unpalatable food are the same in some aspects does not imply that they are the same in all.

A food that induces a learned nausea reaction may well be treated the same as an innately unpalatable food in the ways detailed above. If the purpose of many of the behaviours elicited by innately unpalatable foods is to prevent ingestion, it is not surprising that similar responses are seen to foods previously paired with nausea. As Pelchat et al. (1983) point out, "rats may respond to any diet that they do not intend to eat as they respond to quinine, no matter what the reason" (p.141). Further, that rats respond to foods previously paired with shock differently does not imply that the responses to a food previously paired with nausea and to an innately unpalatable food are identical, as Pelchat et al. (1983) appear to argue. To prove two responses are identical is difficult, one can usually just repeatedly fail to find any differences. As such, the studies mentioned above support, but do not prove, the conjecture. I will come on to the evidence that the responses are different later.

Further evidence for the hedonic shift hypothesis faces a similar problem. DiLorenzo & Garcia (1985) demonstrated that the excitatory responses of a rat’s parabrachial neurons to a palatable taste like saccharin are altered to match the normal responses to an unpalatable taste like quinine by pairing the saccharin with nausea. Chang & Scott (1984) also found neurophysiological changes consistent with a change in the hedonic properties of the food paired with GI malaise. However, we do not know how specific these responses are. Are we simply measuring aversiveness in some more general sense?

Rats and mice make for very good experimental subjects for psychology research. They have been the mainstay for studies on learning for decades. However, for studies on nausea and vomiting, they are a disastrous choice because they cannot vomit. Emesis is our usual marker for nausea, although it is by no means perfect. It is the main way of determining nausea in non-human species. Emesis is seen throughout the vertebrates, but not in every species. Rodents—including rats and mice—are one of the few groups who do not vomit. This idiosyncrasy of one animal group has biased the whole field. No wonder rat research has been uninterested in the possibility of learned nausea responses.

Research on LFAs should have long switched to other animal models. Working with a species that can vomit, cougars (Felis concolor), Gustavson, Kelly, Sweeney & Garcia (1976) paired lithium chloride with bait. The cougars did not show any emesis following ingestion of the lithium, but they did show emesis to subsequent presentations of the bait.

There are obvious logistic difficulties in using cougars in experiments! The interest of Gustavson et al. (1976) was in controlling predation by this species. For more practical studies, shrews and ferrets are increasingly being used by those researching nausea. However, as research has remained concentrated on the rat, we are left trying to interpret this literature.

Although emesis is not found in rats, they do demonstrate pica—the consumption of non-nutritive substances—and this is thought to be a symptom of nausea (Mitchell, Winter & Morisaki, 1977). Mitchell et al. (1977) have shown pica occurs as a response to a flavour previously paired with nausea. Note, similarly, that Rozin (1967) found that thiamine-deficient rats showed "redirected feeding" to their thiamine-deficient diet but not to quinine.

Various other rat studies support a learned nausea hypothesis. Coil, Hankins, Jendin & Garcia (1978) demonstrated that therapeutic doses of four different antiemetics (chosen from each of the four basic categories) attenuated the expression of a LFA. The drugs did not interfere with the conditioned suppression of drinking to a noise previously paired with foot shocks, implying the effect was not a general amnesic one. Neither did the drugs interfere with the rats’ reactions to an innately unpalatable food, again quinine. This would indicate that the reaction to a food previously paired with nausea was one of nausea. However, Goudie, Stolerman, Demellweek & D’Mello (1982) failed to replicate this result. It is worth noting that ANV in cancer patients is notoriously resistant to anti-emetic therapy, suggesting that learned nauseas generally may involve different neural pathways. If a learned nausea from a LFA is similar, then anti-emetics may fail to have any effect.

Vagotomisation interferes with emesis produced by GI stimuli (though not intravenous stimuli) and Kiefer (1985) theorised that the vagal efferents would predominantly mediate any learned nausea. Kiefer, Rusiniak, Garcia & Coil (1981) found that vagotomisation severely attenuates an LFA and the LFA is extinguished faster than for controls.

Reassuring as background, we can note that a considerable body of work has demonstrated various conditioned reactions other than nausea in response to foods. Zahorik (1972) paired flavours with a thiamine deficient diet and with recovery from that diet (through thiamine injections) in rats. Thiamine deficiency produces bradycardia in rats and the flavour paired with thiamine deficiency produced a lower heart rate than the flavour paired with recovery or a novel flavour. Unfortunately, as Zahorik points out, it is difficult to interpret these figures. A novel flavour is not good for producing a baseline measure as rats are neophobic. We cannot conclude whether the flavour paired with deficiency has produced a CR of bradycardia or whether the flavour paired with recovery has produced a tachycardia CR or both.

Various studies have demonstrated conditioned immunomodulation (see reviews by Ader & Cohen, 1985; 1991). In some cases, this has been to food-related conditioned stimuli (Ader & Cohen, 1975 (although, see Kelley, Dantzer, Mormede, Salmon & Aynaud, 1985); Buske-Kirschbaum, Kirschbaum, Stierle, Lehnert & Hellhammer, 1992; Kirschbaum, Jabaaij, Buske-Kirschbaum, Hennig, Blom, Dorst, Bauch, DiPauli, Schmitz, Ballieux & Hellhammer, 1992). The literature in this field has found no need to view conditioned immunomodulation involving such foods as being different from classical conditioning. Bovbjerg and colleagues (Bovbjerg, Ader & Cohen, 1984; Bovbjerg, Kim, Siskind & Weskler, 1987) have suggested a disassociation between the behavioural and immunosuppressive components of the response to a food previously paired with cyclophosphamide, an immunosuppressant that also produces GI malaise, i.e. that the conditioned immunosuppression and the LFA are separate phenomena. What this actually tells us about non-immunological responses—learned or otherwise—is unclear. They do indicate that at least certain physiological responses can be conditioned with food-related conditioned stimuli.

These studies do demonstrate that physiological CRs to foods occur, yet, as we do not know whether a learned nausea in LFAs is a conditioned reaction or how LFAs relate to classical conditioning phenomena, if at all, it is difficult to use these as support for a learned nausea hypothesis.

Often in psychology, when there are two competing theories to explain a phenomenon, both theories turn out to be true. Thus, we could suggest that:

GI malaise following a food produces both a drop in the palatability of that food and a learned response of nausea to that food. Elkins (1991) suggests that both "conditioned gastric reactions (conditioned nausea responses) and negative hedonic shifts (acquired taste dislikes) are probable contributors to emetically induced [LFAs]" (p. 404). Kiefer et al. (1981) also support this idea and, on the basis of their findings about the effects of vagotomies on LFAs (see above), suggest that the purpose of the nausea CR is in maintaining the LFA.

If a dual approach is correct, the question arises of how the two effects are related to each other and whether they are disassociable. The repeated pairings of nausea and food inherent in chemical aversion therapy for alcoholism and in cancer ANV produce the clearest conditioned nausea effects. ANV can occur without the involvement of food-related stimuli. This suggests parallels between the conditioned nausea aspect of LFAs and classical conditioning. Perhaps the LFA phenomenon consists of a drop in palatability plus a more usual classical conditioning with a CR of nausea. This analysis would concur with the conditioned immunomodulation experiments (see above). The research presented herein, however, will concentrate on demonstrating the presence of a learned nausea and is less concerned with tackling whether any direct hedonic shift in palatability is or is not also involved.

LFAs have been interpreted in yet other ways. Weinberg, Smotherman & Levine (1978; Smotherman & Levine, 1978) draw parallels between the neophobia response of rats and their response to a food previously paired with sickness. In humans, Section 2 has described how neophobia can manifest itself in humans either as ‘distaste’, i.e. similar to the reaction to unpalatability, or as ‘disgust’, involving further elements, including nausea. Some papers have talked specifically of a "conditioned disgust". This has been reported in coyotes (Gustavson, Garcia, Hankins & Rusiniak, 1974; Garcia & Hankins, 1977; Gustavson, 1977) in the form of behaviour like urinating on or burying the target food. The extent to which ‘disgust’ as discussed in humans in Section 2 applies to other species is unclear. It seems an unwarranted anthropomorphisation to label these coyote behaviours disgust, however there still remains a behaviour pattern to be explained. What we need to know is whether such behaviour is typical of a coyote to nausea or to unpalatable food or whether this behaviour is unique to LFAs.

Garcia et al. (1985) talk further of disgust: "the critical psychological process is a hedonic shift in the palatability of a taste and/or odor CS, not a CS-US association. The CS has become "disgusting" " (p. 13). It seems here that the authors are simply equating "disgust" with extreme unpalatability as opposed to considering it a qualitatively different reaction, as I have in Section 2. Human ‘disgust’ also includes a nausea response, so with respect to the narrow question that primarily interests us here, perhaps this possible distinction is not too important.

Logue (1979) argues that LFAs are describable under the normal laws of learning connected with classical conditioning, which sometimes produce cathectic changes. Over his many years in the field, Garcia’s views have developed, covering the gamut of possible explanations. In his 1989 review, Garcia talks about cathectic changes and suggests that LFAs produce an affective shift. However, the basis for much of this thesis is that negative affective responses can come in many different forms and palatability/aversion is not unidimensional. LFAs may produce a negative affective response, but that does not define the nature of the aversion.

Furthermore, if we can recognise a variety of different negative affective responses to food within humans, might there also be a variety of different patterns of behaviour in response to the same scenario, a food previously paired with nausea, i.e. rats spilling food; Garcia, Gustavson et al.’s "disgust" reactions in canids etc.

With the animal literature so clouded on the matter, experiments on human LFAs present a great opportunity to try and resolve the matter as humans can be quizzed on their responses to a target food, unlike rats, coyotes or cougars. Unfortunately, the main dependent variable in all the human studies has been the amount consumed of the target beverage or food. This adequately measures the degree of aversion but not the nature of the aversion. Lamon et al. (1977) additionally gave their subjects a taste perception questionnaire, involving ratings on a number of different dimensions (e.g. good-bad, bitter-sweet, thick-thin), the results of which produced no significant differences. Problems with their methodology, though, cast doubt on the significance of their findings. Okifuji & Friedman (1992) additionally measured taste ratings (7 point scale: 1 = excellent to 7 = disgusting). Note how this again confuses the distinction between ‘distaste’ and ‘disgust’ foods by equating disgust with extreme distaste. The effect they found held on both taste ratings and amount consumed. Cannon et al. (1983) had their subjects rate the target beverage on four-point scales for fourteen adjectives (alarming; unpleasant; dangerous; pleasant; relaxing; repulsive; distasteful; appetizing; safe; bad; tasty; harmless; good and harmful). None of these produced significant differences except distasteful, which showed an inverse relationship to amount consumed, being higher following the production of a taste aversion. They also measured heart rate response to their target food but found no effect (Baker et al., 1984). Confusing the matter, Cannon et al. (1980) used apomorphine to induce nausea. This drug’s position in the dichotomy suggested by Hunt & Amit (1987) is unclear. Apomorphine is self-administered by rats, but it does also produce GI malaise. Thus, the LFA induced by Cannon et al. may not be indicative of LFAs induced by GI malaise. This suggests there is some evidence for a drop in palatability (Okifuji & Friedman, 1992; Cannon et al., 1983), but the results are difficult to interpret and none of the studies explicitly considered nausea as a response in the aversion.

The two studies by Rozin & Fallon (1980; Fallon & Rozin, 1983) discussed at length in 2.2 also covered LFAs, albeit with an observational rather than experimental methodology. One of the items for Rozin & Fallon (1980) was "A dislike that you can trace to a particular experience in your past", Ss being asked to name such an item and give details of their experience with it. Analysis of the results then looked at only such aversions that involved GI symptoms, i.e. foods that were previously paired with GI malaise. The various items in Rozin & Fallon (1980) all fitted their hypothesised tripartite categorization well with the exception of such LFA foods and inappropriate breakfast foods ("A food that you strongly like but that you would dislike as your first food of the day"). These fitted the "distaste" category, except for high scores on the statements "The thought of eating this food makes me nauseous" and "I dislike the idea of having this food in my stomach". The second study (Fallon & Rozin, 1983) replicated the result for "aversion" foods, although these where now defined more broadly ("A food that you dislike and for which you can trace the dislike to a particular experience in the past (e.g. being forced to eat it, eating a spoiled sample, getting sick after eating it (whether or not the food caused the sickness), or associating the food with a particularly unpleasant experience.)" Table II, p.18.), while inappropriate breakfast foods now fitted the "distaste" category. Rozin & Fallon (1980; Fallon & Rozin, 1983) interpret these results as fitting their suggested taxonomies. Pelchat and Rozin (Pelchat & Rozin, 1982; Pelchat et al., 1983) have suggested that the differences between a rat’s avoidance of a food previously paired with nausea and its avoidance of a food previously paired with external damage (e.g. shock) parallels the ‘distaste’/’danger’ distinction in studies on human food rejection. However, an alternative interpretation that LFA foods are distinct in the response they evoke seems equally fair from the data in those studies.

With respect to food rejection behaviour, it is not the palatability of a food which is important, but its expected palatability and this is equally so for other characteristics. I reject celery because I expect not to like the taste; I do not re-taste it each time I encounter it, find it unpalatable and stop eating it. With the target item of a LFA, reaction to the anticipation of eating the food need not be the same as the reaction having eaten it. The distinction is impractical to make in animal research, but as we can, to some extent, require our experimental subjects to eat something they find aversive, this is a further potential avenue for human studies.

3.3 Clinical Perspectives

In humans, numerous studies have found nausea and/or emesis to alcohol following aversion therapy (Elkins, 1991, p. 395), i.e. a learned nausea/emesis. However, the aversions induced by such therapies may be atypical of LFAs, see 3.1.2.1. ANV seen in patients undergoing treatments for cancer may be another example of learned nausea if this phenomenon is actually related to LFAs. There is an overwhelming body of anecdotal reports of learned nausea in ANV involving food cues.

I have so far discussed these various learning phenomena involving nausea in the context of what they can tell us about LFAs, but these phenomena are of great clinical importance in themselves. Anticipatory nausea and vomiting (ANV) is a very distressing side-effect of chemotherapy. ANV clearly involves a learned nausea, triggered by a variety of cues, some of which are foods or olfactory.

Learned nauseas may be more widespread. Anecdotal reports exist of a similar process to ANV occurring with pregnancy sickness. Some sort of "conditioned motion sickness" seems plausible and concords with the practical experience of many. The RAF has used desensitising exposure to potential cues of a "conditioned airsickness", while sailors frequently report learned nausea phenomena. If such is commonplace, then it could go some way to explaining the unpredictability of motion sickness.

It is tempting to suggest that cases of nausea and vomiting with no detectable organic, somatic cause may also sometimes be learned nauseas. While the literature remains sparse, it is easy to speculate in this way. These avenues aside, ANV alone merits considerable attention.

ANV may affect up to 50% of patients undergoing chemotherapy and can be an important cause of people dropping out of treatment. Nausea is a very unpleasant symptom, while vomiting causes dehydration and interferes with nutrition. ANV may also contribute further to "cancer anorexia". For unknown reasons, ANV is resistant to anti-emetic therapies, making it especially distressing.

Clinical research on ANV has assumed that LFAs are an appropriate animal model. However, a hedonic shift model of LFAs leaves little room for any connection to the learned nausea of ANV. Demonstrating that LFAs also involve a learned nausea response would affirm a probable link.

Thus, as well as investigating the nature of LFAs, I hope in these studies to demonstrate the production of a learned nausea response in humans in a wholly experimental controlled setting for the first time.

Unfortunately, there is not space within this thesis to consider the next step to take in this research, which would be to investigate the significance of food-related or non-food-related cues. LFAs, by definition, involve food-related cues, yet ANV and other possible learned nausea phenomena usually do not involve food.

3.4 Methodological Issues to do with Nausea in Humans

3.4.1 Generating Nausea in an Ethically Acceptable Way

To produce LFAs in an experimental setting in human volunteers requires that we produce nausea. This raises both methodological and ethical problems. It is difficult to produce nausea in a controlled fashion, nor can we inflict too much discomfort on our experimental participants. Any procedure to induce nausea will probably produce other side effects, adding further complications. These side-effects could be unpleasant or even harmful, adding to problems over the ethical acceptability of a study. Side-effects may also interfere with the psychological or physiological processes we aim to investigate. We clearly want to minimise side-effects, particularly unpleasant ones.

Especially as nausea can be very unpleasant, for informed consent, it is necessary to make subjects fully aware of what they are agreeing to undergo. This requirement may compromise blindness without careful design. Further, there must be an absolute level of discomfort that it would be unacceptable to exceed even with consent. While nausea per se is not dangerous, emesis carries some medical risk and is yet more unpleasant, as well as being inconvenient for the experimenter! It was apparent, for example, that the levels of nausea, vomiting and other symptoms produced by apomorphine in Cannon et al. (1983) would certainly not be acceptable to local ethics committees or the experimenter.

Within these limitations, we still need to produce sufficient nausea for the phenomena being studied, LFAs, to actually occur. Precisely how much nausea is required in humans is unclear, although we can make rough comparisons with the earlier human studies. We know that the strength of an aversion is correlated with the symptoms experienced after the food. Thus, the desire for a strong experimental result involves that we generate symptoms as great as possible within the ethical constraints.

In addition to producing sufficiently strong nausea, we would like the levels of nausea experienced by individuals within the experimental group to be roughly equal. This requires a certain degree of control over the symptoms produced. This is also necessary so as to avoid any subjects suffering ethically unacceptable levels of nausea.

When subjects are re-tested on the target food, they also need to have recovered from the nausea and any other symptoms induced earlier. Given the procedural difficulties in arranging longer studies and in recruiting subjects for such, this means we would like to produce a short-lived nausea to allow the experimental design to be completed over a relatively short time. Also, to inflict a longer-lasting nausea would again raise ethical questions.

Thus, we desire an easily-controlled and fairly transient nausea, which is mild enough to be ethically safe yet severe enough to still produce LFAs and with minimal side-effects

Experimental nausea is usually induced in one of five ways: through imaginal means, motion sickness, pharmacology, X-irradiation or surgery. The latter two are clearly unacceptable here. Human experimental studies on LFAs have used both pharmacological means (e.g. Cannon et al., 1983) and motion sickness (e.g. Arwas et al., 1989), while aversion therapies have also used imaginal processes.

The use of drugs to induce nausea presents a number of difficulties. Individuals’ reactions to drugs may be idiosyncratic and, after administering the drug, there is little way of controlling the resultant symptoms. Symptoms also typically persist for longer than would be convenient for other aspects of the experimental design and side-effects are common. There may also be specific medical dangers associated with the drug.

A task to produce nausea from subjects’ own imagination, perhaps with props of some sort, is clearly a safe method with practically no side-effects. However, it may not produce sufficient levels of nausea and the imaginal procedure may interfere with the design of the psychology experiment.

Motion sickness produces nausea along with a number of other symptoms. It can be produced via a number of procedures. The resulting symptoms fade fairly quickly after the challenge stops. Symptoms experienced correlate with exposure to a motion sickness challenge, allowing some control over the level of symptoms. While motion sickness does produce other symptoms (dizziness and disorientation, sweating and general arousal), these are not clinically dangerous or obviously likely to interfere with the psychological phenomenon of interest. These qualities have made motion sickness a popular way of inducing nausea in past studies. Motion sickness was adopted for the studies in this thesis too.

However, motion sickness does have some disadvantages. Most importantly, individuals’ reactions to the same challenge are enormously variable for reasons still largely unknown. To some extent, because it is simple to ‘titrate the dose’ of motion sickness, less susceptible individuals can simply be given a longer or greater challenge, but the variance of susceptibility is so great that some subjects will not develop symptoms to a particular procedure or do so within a useful timeframe.

Methods to produce motion sickness frequently involve moving the entire subject. Such are expensive and would have been difficult within the logistic constraints of this research. Fortunately, there are some procedures which are considerably simpler. The two used in this thesis are a pseudo-Coriolis effect induced with a circular vection drum; and vestibular stimulation through head movements. The circular vection drum is a rotating cylinder, within which a person is situated. The impression of movement is in conflict with the lack of vestibular response, leading to motion sickness. This procedure was used by Lamon et al. (1977), Okifuji & Friedman (1992) and Hu et al. (1996). To exacerbate the effect, head movements can be made to induce a more strongly conflicting vestibular response (as in Lamon et al., 1977). In the second method, a simple pattern of large head movements can be used to produce a strong vestibular response that will induce motion sickness. The detailed methods used here are described in 3.5.1.1 and 3.5.2.1.

Both also produce the other characteristic symptoms of motion sickness, which were observed throughout the experiments. In both cases, a variable exposure is used such that individuals beginning to experience severe symptoms were stopped, with a control procedure yoked for time. Unfortunately, in both cases, some individuals failed to report more than mild nausea and some reported none at all. This causes problems for the analysis (compare Okifuji & Friedman, 1992), as is considered later (3.5).

3.4.2 Measurement of Nausea

A considerable problem for research in this field is the ineffable nature of nausea. As discussed earlier (3.2.1), without emesis, it is very difficult to measure in rodents. One common way of measuring nausea in these animals, through its ability to produce LFAs, is clearly also rather unhelpful here! That nausea cannot be directly observed is partly why the rat literature tends to use the term "gastrointestinal (GI) malaise", or specifically "upper GI malaise", over the term "nausea" used in human studies.

Even emesis is only a coarse measure and it usually yields only a binary outcome variable. Occurrence of emesis is not even necessarily highly correlated with nausea. Common human experience informs us of vomiting events with little nausea and conversely of considerable nausea without vomiting.

Aside from the obvious fact that participants of our own species can tell us more about ourselves, if we work with human volunteers, we are than able to ask people how they feel and how they feel about a target food. Nausea is usually measured in people through self-report and it is upon self-report that the studies performed in this section rely.

Yet self-report has inherent problems. Is there some alternative? This question leads us back to the question of precisely what is nausea. It seems to be a psychological experience, like pain, as opposed to a physiological state. However, are there physiological correlates of nausea and only nausea?

Electrogastrograms (EGGs) have emerged as a possible source of an answer. EGGs, the use of cutaneous dermal electrodes over the stomach measuring its muscle potential, have been known about for decades (Alvarez, 1922; Abell & Malagelada, 1988) and the concept is readily apparent following the frequent use of various other electromyograms (electro-oculogram, electrocardiogram etc.) and the electroencephalogram.

EGGs can give accurate recordings of stomach muscle activity (Stern, Koch, Stewart & Vasey, 1987) and they reveal a normal pattern of activity entailing a regular cycle of contractions of about 4 cycles per minute (cpm). Crucially for our purposes, nausea seems to involve a changed pattern of stomach muscle activity, namely arrhythmic tachygastria (see 5.3.2 for illustrations). The stomach muscle activity shows an increase in power at both tachygastric and bradygastric frequencies (Stern, Koch, Stewart & Lindblad, 1987). The physiological significance of this is uncertain, but it may relate to a reduced rate of stomach emptying. (The latter is another correlate of nausea but is currently impossible to measure without highly invasive procedures.)

The use of EGGs to measure nausea is still experimental. Studies have yet to show the limits of procedure, nor have they shown a precise relationship with varying levels of nausea. Much of the published literature obscures problems reported by some researchers with obtaining consistent and good recordings and has proved difficult to replicate (e.g. Kingma, 1989). The comparatively mild and transient nausea induced in the experiments herein is less dramatic than the symptoms involved in most of the EGG literature and must be at or beyond the limits of current EGG technology.

EGGs were recorded in the second study, however the results proved disappointing (3.5.2.2.3).

3.5 Two Studies

3.5.1 Study 1

We wish to test the nature of the aversion to the target of a LFA aversion which we have experimentally produced in a controlled design. In particular, I predict that the aversion will involve a learned nausea.

The required experimental design to investigate these issues is straightforward at one level. We need to experimentally generate a LFA to a target item, quizzing subjects on their reactions to that food before and after. Instead of simply measuring the general aversiveness produced, we need to specifically ask about the food’s palatability separately from whether it induces nausea.

The pragmatics of designing such a study are more complicated. In 3.4.1, I discussed the particular problems surrounding generating nausea and conclude that motion sickness is the best tool available. There is also the perennial problem of avoiding response bias from the experimental subjects.

The experiment desribed below and that in the next subsection (3.5.2) both follow the same basic plan. Subjects rate a food item and then either undergo a procedure to induce motion sickness or a control procedure. There is a gap of time before subjects return to rate the food again. Some deception is used to disguise the true purpose of the experiment. As motion sickness is not perfect at inducing sufficient nausea, analyses will take into account the degree of nausea reported.

The study described here was carried out at the same time as the FRI questionnaire study reported in 2.4 seq. The gap between the procedure and the second food tasting was filled by the completion of the FRIs. The methods section below thus overlaps that of 2.4.2.

3.5.1.1 Methods

Design:

A design with unequal sized groups was used so as to ensure sufficient numbers of treatment subjects who felt sufficient nausea as the treatment condition was known from piloting to be only partially successful.

There were two protocols, differing only in the nature of the ‘sensory test’ within the circular vection apparatus: this was either designed to induce motion sickness or was a control procedure.

Subject recruitment:

Subjects were mainly recruited by posters displayed in a number of University of London colleges. The posters stated that "[t]he experiment involves answering a set of questions about various foods, a few of which will also be tasted. The experiment also involves a short sensory test that may induce mild "motion sickness"." Some subjects were contacted directly from a subject panel or were recruited by word-of-mouth. When prospective subjects were first spoken with, the details of the experiment were reiterated along the same lines as the poster description. The warning as to the symptoms that the "short sensory test" could provoke did not stress nausea, but mentioned it among other symptoms (dizziness, disorientation, sweating, eye strain etc.). Any specific questions about the study asked were answered so as not to reveal the exact purpose behind the study or draw attention to nausea. The precise ‘foods’ to be tasted were detailed. Subjects were asked to eat a normal lunch beforehand, avoiding alcohol.

Exclusion criteria:

Subjects were excluded with any history of epilepsy or of ‘classic’ migraine attacks, involving either nausea or visual disturbance. Other exclusion criteria were concurrent GI disorders or concurrent medication indicative of abnormal psychology or GI disorders and previous experience of the test food, kśrrta (see below). A minimum age of sixteen was imposed, but there was no maximum age.

Protocol:

All experimental sessions took place in the afternoon (Monday to Saturday), from October 1993 to February 1994. Sessions started around 2.30 p.m.

Subjects were met and taken to room A. The course of the afternoon was explained and basic demographic questions asked. The subject was then taken to room B, where the circular vection apparatus was housed, and the details of the "sensory test" were described: namely, that its purpose was to measure the "subject’s sensitivity to dizziness", that it may produce a number of symptoms (which were listed) and that the subject could and should terminate the procedure at any time if they felt too ill.

The different methods for answering questions was explained with examples: dichotomous (e.g. True/False), Likert scales and visual analogue scales (VASs). It was then explained that, as many people were unfamiliar with VASs, all subjects were to practise their use. For this "practice", a sample of kśrrta (the target food) would be used. It was explained that kśrrta would be re-encountered "properly" later on. Subjects then completed three VASs for their sample of kśrrta: one and three of which were as FRI questions 2 and 8; practice question two was "How familiar is kśrrta to you?" answered on a VAS with anchors of not at all familiar to very familiar and no midpoint.

The subject answered further questions: a VAS hunger rating, state STAI and a questionnaire about previous experiences of disorientation, dizziness, nausea and vomiting. If a subject indicated "4 to 10 times" or "More than 10 times" for any symptoms on this, the circumstances of these events were determined. The exact procedure for the circular vection apparatus was then explained and the subject sat in the apparatus.

Subjects underwent either a procedure to induce motion sickness or a control procedure in the circular vection apparatus. Control subjects were yoked to experimental subjects for time spent in the apparatus. All subjects rated how they felt on four symptoms (disorientation, dizziness, nausea and sweating) on a 5-point scale: 0 = None. Symptom not present; 1 = Slight; 2 = Medium; 3 = Considerable; 4 = Extreme. The subject had been familiarised with this scale before entering the apparatus.

After the subject was positioned in the apparatus, they made their baseline symptom ratings. The circular vection machine was then started, subjects being asked for symptom ratings every two minutes, immediately after starting or stopping directed head movements. The last rating (after ten minutes) was done just after stopping the machine. The subject then sat outside the apparatus for a further two minutes rest, followed by another round of symptom ratings. (This rest period and symptom rating was omitted if the subject had given ratings of 0 at all previous stages.) Unless the subject wished to rest longer, they and the experimenter returned to room A.

If subjects gave symptom ratings of 4, they were reminded that they could stop the procedure. If subjects did stop the procedure, they were brought out of the apparatus, the time noted and symptom ratings taken. They were then given a two minute rest, and the experiment continued as for other subjects.

In room A, symptom ratings were again asked, the time being noted. Water was available ad libitum from now on. If there were any non-zero symptom ratings then, the subject was left for a few more minutes; after which, symptom ratings were again asked. The subjects completed several paper-and-pencil measures. If the last symptom rating had produced any non-zero responses, another round of ratings was then made. Subjects then completed multiple FRIs and the experiment continued as detailed in 2.4.2.

At the end of the FRIs, subjects tasted a number of items. The first of these was the test food. Subjects were presented with a kśrrta sample and filled in a FRI before tasting it. They then tasted it and filled in a second FRI. They then proceeded to taste the other items and the experiment finished as detailed in 2.4.2.

Materials—Circular vection apparatus:

Circular vection to induce a pseudo-Coriolis effect was used. For the circular vection apparatus, also known as an optokinetic drum, subjects are seated in a rotating cylinder, giving the impression of movement and causing motion sickness. Here, subjects sat surrounded by a cylinder of curtain fabric (26 inch diameter), with vertical black and white stripes, filling their field of vision.

For experimental subjects, the curtain rotated at 10.5 cpm. Subjects rocked their heads from side to side; the experimenter having demonstrated this action at about 1 Hz. For control subjects, the curtain was stationary and they nodded their heads up and down. For both groups, head movements were performed in alternate two minute periods, with head movements during the first interval. Subjects were periodically observed from above the apparatus to check compliance with their instructions.

Materials—Kśrrta:

Various factors determined the choice of the aversion target food item. Familiarity protects against the formation of LFAs (Cannon et al., 1983; Arwas et al., 1989), so a novel substance was required. The best way to guarantee novelty was to invent a new substance. This also reduced the ethical worry that subjects could be left with an aversion to a safe food they might otherwise have been able to enjoy. Previous human LFA studies (e.g. Cannon et al., 1983; Arwas et al., 1989, Lamon et al., 1977) have all used beverages as their target items, as have most non-human studies. As most foods are solids not liquids, the use of a solid food item was felt to be more naturalistic.

Kśrrta was developed by myself and Alex Westcombe as a novel food item for use in a number of studies. It was first used as reported in Westcombe & Wardle (1997), where it proved to be a most useful recipe. Westcombe & Wardle (1997) found it to be highly novel. It was felt sensible to avoid using something at the extremes of palatability to begin with and Westcombe (pers. comm.) found kśrrta to have a mean liking rating of 0.0 on a -4 to +4 Likert scale.

Kśrrta was prepared as in Westcombe & Wardle (1997). The recipe consists of a 3:1 ratio (by weight) of puréed prunes (prepared from tinned prunes, drained) to tofu. Batches were produced en masse and then frozen. Samples for each subject were defrosted on the day of use.

Materials—Questionnaires:

Described in 2.4.

3.5.1.2 Results

One subject was excluded because of suspected psychopathology. Due to an error, one subject used had previously encountered kśrrta in another experiment (that described in Westcombe & Wardle, 1997) and was therefore excluded from analyses concerning kśrrta. There were 21 subjects of each sex. Ages ranged from 20 to 51, with a median of 24.5. The modal age was twenty with 10 subjects.

The various background variables were explained and considered as part of Section 2 (2.5.1).

3.5.1.2.1 Symptom Ratings

Past symptomatology:

Subjects filled out a symptom history questionnaire, covering 4 symptoms (disorientation, dizziness, nausea, vomiting) over the past 6 months (Table 3.1).
 

 

Symptom

Never Once 2 or 3 times 4 to 10 times More than 10 times Low median
Disorientation
21
7
7
5
2
Never
Dizziness
11
10
10
10
1
Once
Nausea
10
9
18
5
0
2 or 3 times
Vomiting
22
13
6
1
0
Never
Table 3.1: Symptom history.

If a subject checked 4 to 10 times or More than 10 times, they were asked for further details of these events. Reasons given were highly variable: alcohol was one of the few repeated; others included playing wind instruments and a visit to an amusement park.

Symptom ratings during the study:

Only 3 experimental subjects did not complete the full ten minutes in the apparatus, lasting 2:58, 3:01 and 3:04 minutes respectively. The treatment procedure was halted after the same times for the yoked control subject.

At baseline, most subjects responded 0 for all the symptoms (Table 3.2).
 

Symptom

0 None, not at all 1 Slight 2 Medium
Disorientation
31
11
0
Dizziness
37
4
1
Nausea
42
0
0
Sweating
39
3
0
Table 3.2: Symptoms at baseline.

Peak symptom ratings were calculated for each subject. Table 3.3 shows intercorrelations between the peak symptom measures for those subjects who underwent circular vection. All results are one-tailed.
 

Variables

Peak disorientation Peak nausea Peak sweating
Peak dizziness rS = 0.58;
p < 0.001
rS = 0.38;
p = 0.017
rS = 0.35;
p = 0.026
Peak disorientation   rS = 0.28;
p = 0.06
rS = 0.39;
p = 0.015
Peak nausea   rS = 0.49;
p = 0.002
Table 3.3: Relationship between peak symptom ratings; ns = 32.

If we plot the symptom ratings against time for each subject, an integral, or area under the curve, measure can also be calculated (using straight line interpolations between points) as a measure of the total amount of that symptom experienced; thus equating a subject who, say, felt very nauseous for a brief period of time with one who felt mildly nauseous for a longer time. This was done for the key nausea variable only. Integral (total) nausea was very highly correlated with peak nausea (rS = 0.95, p < 0.0001, one-tailed). This suggests that there is little to be gained from using the integral score in addition to the peak value.

There were two procedures: circular vection (CV; n = 31) and the control procedure (C; n = 13). The results are displayed in Table 3.4.
 

 

Median (IQR)

None, not at all Slight Medium Considerable Extreme
Symptom
CV
C
CV
C
CV
C
CV
C
CV
C
CV
C
Peak dis-orientation
2

(1.5)

1

(2)

3
4
5
3
11
2
10
1
3
0
Peak dizziness
3

(2.5)

1

(2)

3
4
5
2
7
2
7
1
10
1
Peak nausea
1

(2)

0

(0)

15
8
7
1
5
1
5
0
0
0
Peak sweating
1

(1.5)

0.5

(1)

14
5
10
3
4
2
4
0
0
0
Table 3.4: Peak symptom ratings by procedural group.

Using Mann-Whitney tests, there are significant differences for peak disorientation (z = 2.63, p = 0.008), peak dizziness (z = 2.25, p = 0.024) and marginally so for peak nausea (z = 1.85, p = 0.064), but not for peak sweating (z = 0.51, p = 0.6). As intended, circular vection produced motion sickness.

Of the 32 subjects who underwent the treatment procedure, only 10 reported a peak nausea of 2 or more. These form the experimental group, with the other 22 forming a failed treatment group; this is following the approach taken by Arwas et al. (1989). One control subject gave a peak nausea rating of 2 and is excluded from subsequent analyses. The results for the peak symptom scores for the experimental (Ex) and failed treatment (FT) groups are displayed in Table 3.5.
 

 

Median (IQR)

None, not at all Slight Medium Considerable Extreme
Symptom
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Peak dis-orientation
3

(1)

2

(2)

0
3
1
4
3
8
4
6
2
1
Peak dizziness
4

(1)

2

(2)

0
3
0
5
2
5
2
5
6
4
Peak nausea
2.5

(1)

0

(1)

0
15
0
7
5
0
5
0
0
0
Peak sweating
1

(1)

0

(1)

1
13
5
5
2
2
2
2
0
0
Table 3.5: Peak symptom ratings in the circular vection group, split on peak nausea ratings.

Unsurprisingly, by splitting on peak nausea, we also see differences on the other variables.

3.5.1.2.2 Responses to Kśrrta

Four batches of kśrrta were made, consumed respectively by 9, 20, 11 and 2 subjects. As quality control, a MANOVA was performed on the time 1 kśrrta palatability and nausea ratings against batch. Data was only available for one person in the last batch, so only the first three batches were considered. There was no effect of kśrrta batch (ps > 0.1).
 

Variable (101mm VAS)

Median IQR Minimum Maximum
"Taste" (large numbers mean less liked)
50
30
4
85
"Nauseating" (large numbers mean more nauseating)
2
6
0
69
"Familiarity" (large numbers mean more familiar)
2
5
0
99
Table 3.6: Time 1 kśrrta ratings (all n = 41).

The nauseating and familiarity ratings show floor effects. For both ratings, 33 subjects scored below 10. Nearly all the subjects rated the kśrrta as unfamiliar (ten subjects scored 0), but with occasional responses being higher. Even then, it seemed that those subjects who rated their familiarity with kśrrta as high (seven scored over 50) had interpreted the question in a fundamentally different way to other subjects, more whether the taste is like any other taste. Thus, we can be satisfied that kśrrta was suitably novel.

For the initial nausea ratings for kśrrta—that is, how much nausea did kśrrta produce—fourteen subjects gave a zero rating and all but one gave ratings below 25. The one other rating was 69 and the subject with this score volunteered the information that this was because kśrrta reminded him of sesame, to which he appeared to have an LFA. Consequentially, this subject’s data was excluded.

Kśrrta and hunger ratings were on VASs and were transformed using an arcsine root function (see 2.5.2). For subsequent analyses, the transformed scores will be used.

Comparing the initial kśrrta ratings of palatability and nauseatingness to each other, there was a weak correlation: r = 0.30, p = 0.06.

At the end of the experiment, the kśrrta was rated twice with a set of FRI questions: once before tasting and once after tasting. The two ratings were very closely related: the first canonical correlation between all 13 scores for each is greater than 0.99. The minimum individual correlation between each question was 0.51 for Q9. The after tasting scores are the primary measure and analyses of the before tasting scores will only follow if effects are found with the after tasting scores.

Difference scores were calculated for the palatability and nauseatingness (transformed) scores of the after tasting scores minus the initial scores. A MANOVA was performed with a within-subject factor of the two kśrrta difference ratings and a between-subjects factor for the two groups (Experimental and Control; ignoring Failed Treatment). There is no significant effect of group: F2, 16 < 1, p = 0.5. Means (standard deviations) are given in Table 3.7:
 

Variable
Ex
FT
C
Taste [high means became less palatable]
-0.063 (0.305)
0.053 (0.292)
0.052 (0.314)
Nauseating [high means became more nauseating]
-0.028 (0.145)
0.067 (0.208)
0.037
(0.115)
n
10
19
9
Table 3.7: Change in kśrrta ratings by group.

Means (standard deviations) for the transformed scores at the two times are in Table 3.8.
 
Pre-procedure scores After tasting scores
Variable
Ex
FT
C
Ex
FT
C
Taste
[high is less palatable]
0.80 (0.17)
0.81 (0.21)
0.68
(0.28)
0.74 (0.32)
0.88 (0.42)
0.73 (0.48)
Nauseating
[high is more nauseating]
0.27 (0.20)
0.11 (0.14)
0.13 (0.09)
0.24 (0.27)
0.18 (0.23)
0.16 (0.11)
n
10
20
9
10
19
9
Table 3.8: Time 1 and 2 kśrrta ratings by group.

There is minimal change of any sort.

3.5.1.3 Discussion

Given the vast amount of evidence concerning LFAs and their formation, it would be foolhardy to interpret the negative results reported here as evidence against current theories. This leaves the question as to why LFAs where not observed here. There are a number of possibilities.

It is very possible that the nausea produced was insufficient. No subjects gave peak nausea ratings of 4; only five did of 3. As the degree of aversion produced is known to be proportional to the degree of nausea involved, the low levels of nausea involved here may have had only small or no effects. It is difficult to compare levels of nausea produced across different studies with objective measures of nausea unavailable, however clinical studies have involved the much higher levels of nausea produced by either cancer chemotherapy or by emetic aversion therapy, while one of the four experimental studies in humans (Cannon et al., 1983) used apomorphine and reported most subjects vomiting.

The initial amount of kśrrta consumed was generally very small, typically only a few millilitres volume, as subjects were left to taste as much or as little as they wanted. This may have been insufficient for LFA formation.

The time courses involved—between initially tasting kśrrta and the circular vection procedure and between then and subsequent ratings—fall well within intervals used in other experiments and so should not be responsible for the lack of result. Nor is there any obvious reason why kśrrta should be particularly immune to LFA formation or particularly prone to any other effects to mask LFA formation. Initial palatability ratings were normally distributed around neutrality, while initial nausea ratings were at floor—typical of the pattern expected of a novel, plant-based food (see 2.8.1).

Given the lack of evidence for LFA formation, I am unable to tackle the question I sought to answer.

3.5.2 Study 2

In order to investigate the nature of LFAs, we need to reliably produce them. This should be possible in people within an experimental setting using motion sickness, as has been achieved in previous studies reviewed earlier (see 3.1.1.2, 3.4 and 3.5.1.1). We cannot say for certain why the previous study failed to produce detectable LFAs, but an obvious factor was the levels of nausea produced. Thus, it was decided to repeat the basic design of this component of the study using a more nauseating procedure and incorporating some other improvements.

A number of ways of producing greater motion sickness were considered. A method more efficacious than before was required, yet, on ethical grounds, I still wanted to avoid producing extreme symptoms by the use of an easily controlled procedure, ruling out drug treatments and certain methods of motion sickness induction. The use of dramatic head movements, causing vestibular stimulation, was chosen. This is a more direct way of inducing motion sickness than the sensory conflict of the pseudo-Coriolis effect and should be expected to produce greater symptoms (Michael Gresty, pers. comm.).

The use of kśrrta was also reviewed. In the previous study, subjects had simply to taste a small amount of the food and could get away with consuming very little. In the following study, subjects were instead made to eat a small pot of the food. The kśrrta recipe was altered to make it more palatable to begin with, by adding sucrose, to avoid floor effects in subjects’ ratings. Kśrrta is a very bland food, as many subjects in the previous and other studies will testify. The salience of a stimulus is usually associated with learning behaviour, so I attempted to make kśrrta a more salient food through the addition of a distinctive, yet largely novel, odour by adding orange flower water as well. This latter goal may counteract the first, for a more salient novel food may evoke greater neophobia, but piloting suggested the new recipe served its purpose.

Given the importance of nausea, specifically how much was produced through the experimenters’ attempts to induce motion sickness, it is unfortunate that the measurement of nausea levels has been restricted to self-report. Thus, the electrogastrogram (EGG) and recordings of pulse rates were introduced into the experimental design as well.

The last study had been part of a larger process using FRIs to investigate food rejection behaviour, as covered in Section 2. It was not thought that that process would seriously interfere with the LFA component, yet that remains a possibility and the study presented here was carried out alone.

These changes aside, the basic format remains the same as in the previous study and as in numerous other studies on humans or rodents. Subjects taste a novel food, have nausea induced (or not, in a control group) and then, some time later, re-taste the target food. Background measures are similar to those used before, but have been refined following the experiences of that study and in the light of more recent work.

3.5.2.1 Methods

Design:

A design with unequal sized groups was used so as to ensure sufficient numbers of treatment subjects who felt sufficient nausea.

There were two protocols, differing only in the nature of the ‘balance test’: either a vestibular stimulation (to provoke nausea) or a control procedure.

Subject recruitment:

Subjects were recruited through posters displayed in colleges and hospitals (on staff noticeboards) local to the Institute of Neurology and through word of mouth. The posters mentioned the involvement of a "‘balance test’" and the taking of physiological measurements. When prospective subjects contacted the experimenter, the experiment was explained as involving "various questionnaires", "some physiological measures, involving sensors attached over your stomach", "having to taste a food" and a "balance test: a series of bodily movements that may provoke dizziness, disorientation and other symptoms of motion sickness". The food was described as having been specially designed for this study. (Subjects with dietary restrictions were told more about the food, e.g. that it was vegan.)

Subjects booked for near lunchtime were asked to eat an early lunch that day. All subjects were asked to not consume any alcohol that day.

Exclusion criteria:

Subjects were excluded if they had concurrent GI disorders or severe ill health of any kind. A minimum age of sixteen was imposed, but there was no maximum age.

Protocol:

The study was carried out at the MRC Human Movement and Balance Unit, Institute of Neurology, from February to June 1995, with most earlier in this period.

If subjects queried the purpose of the experiment (either before or during the experiment), they were told that they were acting as control subjects for a study on the effect of chronic middle ear infections in childhood. Such ear infections, it was explained, can lead to balance problems and, through affecting the nerves running from the tongue, the sense of taste. (While not the purpose of this study, childhood ear infections do appear to have this effect (Bartoshuk & Duffy, 1994).)

Subjects came in for two sessions. In the first session, the course of the experiment was explained and all subjects signed a consent form detailing the more disagreeable or intrusive parts of the experiment (the ‘balance test’ and the placement of the EGG electrodes). Various demographic and medical history questions were asked. So as to allow time for the signal to settle before any measurements, the EGG electrodes were then attached. Subjects then completed a symptom history questionnaire and the MSQ, followed by the other psychometric questionnaires, introduced by a covering sheet explaining how to answer them. Finally, subjects completed the time 1 state STAI and hunger rating.

The symptom ratings were then explained and subjects gave their first ratings before the first recording block of physiological recordings. Each set of recording block took five minutes, during which time the subject was requested to remain silent. The pulse sensor was attached just before the recordings started. Next, subjects consumed all of a kśrrta sample and rated it on the scales provided (all on one A4 sheet). No mention was made of the second tasting to come.

The ‘balance test’ protocol was explained, with the subject practising a few head or arm revolutions to check compliance with the instructions. The potential side-effects and that subjects were free to stop at any point were re-iterated. It was stressed that the aim was not to make them feel "very ill". It was also explained that the experimenter might independently stop the procedure. The same warnings were given to subjects whichever protocol they were to undergo. Immediately prior to starting the ‘balance test’, subjects completed the time 2 state STAI and hunger rating, then gave symptom ratings.

Both procedures took place in two minute blocks. Subjects gave symptom ratings at the end of each block; with the next block starting promptly afterwards. As symptoms can dissipate very quickly with the vestibular stimulation, all subjects were asked to give a rating for how they had felt during the two minutes. Subjects were allowed to stop during a block, but none did. As soon as a subject stopped whichever procedure, the second physiological recording block occurred.

Subjects then left for ninety minutes. They were supplied with a form so as to complete their symptom ratings at thirty and sixty minutes after leaving. During this time, the EGG electrodes remained attached and subjects were asked not to eat anything or to drink any caffeinated beverages, although other drinks (such as fruit juice) were allowed.

Immediately on returning for the second session, subjects gave symptom ratings. The course of the second session was re-explained. Symptom ratings were again taken, then another physiological recording block, then symptom ratings again and subjects filled out the time 3 state STAI and hunger rating. Subjects were given another sample of kśrrta to consume and rate as before. More symptom ratings were followed by the final physiological recording block, followed by the final symptom ratings.

Lastly, subjects were debriefed, paid and the experiment explained. Subjects were asked not to describe the details of the experiment to other potential subjects. No subject complained of the deception that had occurred or about any other aspects of the experiment. However, it was clear that both the vestibular stimulation and control procedures could be unpleasant and that a few female subjects were embarrassed with the placement of EGG electrodes. That most people believe their stomach to be in their abdomen caused some female subjects to be unprepared for the EGG electrodes’ actual positioning.

ProtocolVestibular stimulation:

For the vestibular stimulation, subjects made large and quick circular movements of the head, with their eyes open, as though looking at the numbers on a clock in turn. The experimenter demonstrated this procedure at approximately 1Hz. For the control procedure, subjects made circular movements of their outstretched right arm, with their eyes shut. The experimenter also demonstrated this procedure at approximately 1Hz, with a radius (at the hand) of about 20 cm. Control subjects performed the procedure for a yoked period of time.

MaterialsKśrrta:

To make this revised recipe for kśrrta, for 100 g of tofu (drained), add 33.3 g of tinned prunes (drained, stoned, puréed and sieved), 5.6 g of water, 2.8 g of sucrose and 11.8 g of orange flower water. All the ingredients were mixed in a food processor. Kśrrta was originally designed for two earlier studies (see 3.5.1.1).

The kśrrta was made in batches and frozen. Samples were defrosted (for serving at room temperature) on the morning of the experiment. Each sample contained about 14g and was presented in a small container. Random samples from each batch were weighed. On a one-way ANOVA, there was no significant difference in mass by batch: F4, 48 = 1.94, p > 0.1. The overall mean (standard deviation) for the samples measured was 13.9g (1.1g).

MaterialsQuestionnaires:

Subjects completed a symptom history questionnaire, trait Spielberger STAI (Spielberger et al., 1983), Absorption subscale (Tellegen, 1978/1982; 1982; 1992), modified ‘Cincinnati Neophobia Scale’ (CNS; see below), 13-item MCSD (Reynolds, 1982) and state STAI (Spielberger et al., 1983) sequentially. The CNS was a simplified version of the neophobia questionnaire devised and used by Raudenbush et al. (1995), but subjects merely answered Yes or No as opposed to the original Likert scale: see 5.3.1.

Kśrrta was rated on seven different questions, on 99mm visual analogue scales (VASs). Hunger ratings were also made on a 99mm VAS.

Symptom ratings were given on a 5-point Likert scale: 0 = Never. Not at all; 1 = Slight; 2 = Medium; 3 = Considerable; 4 = Extreme. Subjects answered for four symptoms: disorientation/dizziness, nausea, sweating and general malaise (described also as "generally feeling unwell"). For the symptom history questionnaire, subjects answered how often they had experienced each of four symptoms (disorientation/dizziness, nausea, vomiting, ear ache) in the last six months on a scale of Never, Once, Two or three times, Four to ten times, and More than ten times. If a subject answered Four to ten times or More than ten times, they were asked about the cause of these events.

MaterialsPhysiological measurements:

All physiological recordings were stored on VHS videotape (recorded on a Racal V-Storer at 15/8 speed) and analysed on a Schlumberger SI 1220 FFT Spectrum Analyzer. Spectral analyses were performed on both the pulse and EGG recordings, using a 50Hz baseband with the tape playing at 8 times normal speed, giving a resolution of 12.5mHz (0.75 cpm). All EGGs were printed out for visual analysis; examples are in 5.3.2. About five minutes of clean recording will provide 3 or 4 periods over which the spectral analyses were averaged. The longest continuous clean period for the signals was analysed. (If there were two periods of about equal length, the earlier one was used.) For the pulse signal, at least 2 averages were so obtained. There were greater problems with the EGG signal. Through drift or sudden movement by the subject (coughing etc.), the EGG signal would overload and have to be re-zeroed. Thus only 1 average was often obtained for the EGG (see 3.5.2.2.3 and 5.3.2).

Pulse was recorded with a pressure transducer attached to the left index finger. The modal frequency of the pulse signal (above 0.2Hz) was noted. If two modes existed, the smaller was taken.

For the EGG analyses, an a.c. coupling (Electronic Designs Limited EDL 2896) with a time constant of 8 seconds was used to filter out very low frequency drift (compare Stern, Koch, Stewart & Lindblad, 1987). The percentage of total power at normal frequencies (2.25-3.75 cpm) and at tachygastric frequencies (4.5-9 cpm) was recorded (compare Koch, Xu, Bingaman, Stern, Demers, Seaton & Summy-Long, 1994; Chen, Schirmer & McCallum, 1994; Xu, Koch, Summy-Long, Stern, Seaton, Harrison, Demers & Bingaman, 1993; Abell & Malagelada, 1988). High frequency contamination from the cardiac pulse is thus avoided.

3.5.2.2 Results

There were 47 subjects. Three subjects undergoing the vestibular stimulation protocol were unable to complete the vestibular stimulation task satisfactorily and only data from before the vestibular stimulation is included. This left 14 experimental subjects who gave a peak nausea score of 2 or above, 13 matched controls and 17 subjects who underwent the experimental protocol but who reported a peak nausea score of less than 2 (failed treatment).

There were 17 men and 30 women. 32 subjects were students (16 postgraduates), 7 were unemployed, 6 worked as science research assistants (at a pre-doctoral level) and 2 held secretarial jobs. Of the 39 who were attached to an educational establishment, 12 were at UCL and 5 or less from thirteen other colleges (with only 3 subjects from establishments outside the University of London). A simple ‘knowledge’ score was constructed to indicate the expected amount of knowledge a subject might have about the nature of the experiment: this was based on their occupation and conversation with the subject. Psychologists (or those who had worked in the field) score 3 (4 people); those in medical or behavioural fields score 2 (14 people); with all others scoring 1 (29 people). In debriefing, no subject knew the purpose of the experiment and no subject admitted to having eaten or to having drank caffeinated products between the two sessions.

Ages ranged from 18 to 64, with a median of 24. The modal age was nineteen (with 18 subjects).

Subjects were booked to start at 13:00, 14:00 or 15:00. The first noted time point was at the beginning of the first physiological recording session, the median time for which was 14:30 (range 13:17 to 16:00). The median time at which the experiment finished was 16:53 (n = 40; range 15:39 to 18:23). The amount of time since the subject last ate something (usually lunch) ranged from 29 minutes to 17 hours, 28 minutes; mean (standard deviation) = 2:49 (2:40).

18 subjects reported dietary restrictions. This included 2 vegans, 6 ovo-lacto-vegetarians, 4 who ate fish but no meat and 1 subject who ate no ‘red’ meat. 34 of the subjects did not smoke. There were 5 regular smokers and 7 occasional smokers (with one missing data point).

Details of the various psychometric measures are given in Table 3.9.
 

Variable

Mean Standard deviation Minimum Maximum n
Trait STAI
42.4
11.8
20
70
46
State STAI: time 1
33.8
8.1
20
62
46
State STAI: time 2
35.1
8.6
20
53
45
State STAI: time 3
32.9
9.0
20
57
44
Absorption
21.7
6.8
8
34
47
CNS (log transformed)
1.0
0.8
0
2.5
46
MCSD
5.6
2.4
0
11
46
Hunger ratings (arcsine root transformed)
Hunger time 1
0.61
0.30
0
1.32
46
Hunger time 2
0.53
0.33
0
1.32
45
Hunger time 3
0.88
0.38
0
1.57
43
Table 3.9: Descriptive statistics for background variables.

Hunger ratings were on 99mm VASs that were transformed following the pattern in Section 2. The CNS showed a floor effect, with 13 subjects scoring 0 (the mode), and a skewed distribution (skew = 1.3). A log transformation improved normality (skew = 0.2). Raw scores for these variables are given in Table 3.10.
 

Variable
Median
IQR
Minimum Maximum
CNS
2
3.25
0
11
Hunger time 1
32.5
35
0
93
Hunger time 2
20
46
0
93
Hunger time 3
69
47
0
99
Table 3.10: Descriptive statistics for CNS and hunger ratings, untransformed.

Responses to kśrrta:

Five batches of kśrrta were made, consumed respectively by 8, 11, 13, 7 and 8 subjects. As quality control, a MANOVA was performed on the time 1 kśrrta ratings against kśrrta batch. There was no effect of kśrrta batch (ps > 0.1).
 

Variable
(99mm VAS)
Median
IQR
Minimum Maximum
Taste (large numbers mean more liked)
25
38
0
86
Creamy (large numbers mean more creamy)
48
51
0
96
Contamination (large numbers mean more mindful of contamination)
17
45
0
91
Sweet (large numbers mean more sweet)
54
40
1
92
Nauseating (large numbers mean more nauseating)
3
12
0
69
Sour (large numbers mean more sour)
6
30
0
65
How hungry (large numbers mean more hungry before subject would eat it)
66
40
0
99
Table 3.11: Time 1 kśrrta ratings (all n = 47).

 

The nauseating and sour ratings show floor effects. For the nausea rating, 33 subjects scored below 10; for sour, 24 subjects. As in Section 2, these VAS scores were transformed using the arcsine root function.

Comparing the initial kśrrta ratings to each other, we find some significant results, shown in Table 3.12.
 

Variable

Nauseating How hungry
Taste r = -0.38;
p = 0.009
r = -0.65;
p < 0.001
Nauseating   r = 0.32;
p = 0.030
Table 3.12: Relationship between time 1 kśrrta ratings.

Taste, How Hungry and Nauseating are correlated with each other, but Contamination showed no significant relationships.

3.5.2.2.1 Symptom Ratings

Subjects filled out a symptom history questionnaire, covering 4 symptoms (disorientation/dizziness, ear ache, nausea, vomiting) over the past 6 months. The full results are given in Table 3.13.
 

Variable

Never Once 2 or 3 times 4 to 10 times More than 10 times Median
Disorientation/ dizziness
23
7
9
6
2
Once
Ear ache
28
13
5
0
1
Never
Nausea
15
9
16
5
2
Once
Vomiting
26
14
3
3
1
Never
Table 3.13: Symptom history data.

If a subject checked 4 to 10 times or More than 10 times, they were asked for the cause of these incidences. These responses were varied, from epilepsy to doing aikido, with alcohol the only recurrent cause. These figures are typical for a normal population (Dr Michael Gresty, pers. comm.).

These four variables were compared with each other with Spearman’s correlations. Past nausea and past vomiting were correlated (rS = 0.25, p = 0.048, one-tailed). Past nausea was also related to past disorientation/dizziness (rS = 0.32, p = 0.030) and past ear ache (rS = 0.36, p = 0.014).

At baseline, most subjects responded 0 for all the symptoms, with Sweating responses being the main exception.
 

Variable
0 None, not at all 1 Slight 2 Medium
Disorientation/ Dizziness
38
8
1
Nausea
40
7
0
Sweating
32
14
1
General malaise
39
7
1
Table 3.14: Symptoms at baseline.

Before the ratings immediately prior to the vestibular stimulation or control task (state STAI, hunger, symptom ratings), subjects’ experiences have been almost identical. The only difference is in the different descriptions of the nature of the "balance test". It may be apparent from the descriptions that the control task is not likely to cause severe motion sickness (although equal warnings were given to each group about the possible "side-effects"). Some control subjects did seem surprised at the mildness of the task described to them.

The time 1 and time 2 state STAI scores were highly correlated (r = 0.83, p < 0.001, n = 44) and not significantly different on a paired t-test (t43 = 1.46, p > 0.1).

The time 1 and 2 hunger ratings are very close together, although subjects do consume the kśrrta in between, which may cause a difference. The time 1 and time 2 hunger ratings were highly correlated (r = 0.84, p < 0.001, n = 44) and the time 2 scores were significantly lower on a paired t-test (t43 = 2.59, p = 0.013). The time 3 rating is much later, so we should expect the subjects to be rather more hungry. The time 3 ratings was significantly correlated with both the time 1 (r = 0.68, p < 0.001, n = 42) and time 2 (r = 0.62, p < 0.001, n = 41) ratings and significantly higher than both (t41 = 6.05, p < 0.001; t40 = 6.05, p < 0.001 respectively)

The symptom ratings immediately prior to the vestibular stimulation or control task, like the baseline ratings, were mainly 0.
 

Variable
0 None, not at all 1 Slight 2 Medium
Disorientation/ Dizziness
39
8
0
Nausea
39
6
2
Sweating
34
11
2
General malaise
39
6
2
Table 3.15: Symptoms immediately prior to vestibular stimulation or control.

There were two treatment procedures: vestibular stimulation (VS; n = 31) and the control (C; n = 13) procedure. The peak symptom score is the peak from between after 2 minutes of treatment to the second in between rating. The results are displayed in Table 3.16.
 

 


Mean (standard deviation)
None, not at all Slight Medium Considerable Extreme
Variable
VS
C
VS
C
VS
C
VS
C
VS
C
VS
C
Peak dis-orientation/ dizziness
2.7 (1.1)
0.8 (1.0)
1
6
3
4
9
2
10
1
8
0
Peak nausea
1.4 (1.1)
0.8 (0.9)
8
7
9
2
9
4
4
0
1
0
Peak sweating
1.8 (0.9)
0.9 (0.8)
2
4
8
6
16
3
4
0
1
0
Peak general malaise
1.4 (1.1)
1.4 (1.1)
6
3
12
5
9
2
2
3
2
0
Table 3.16: Peak symptom ratings by procedural group.

Comparing between the two treatment procedures with a MANOVA, with symptom as a within-subjects factor, there is a significant treatment procedure by symptom interaction (F3, 40 = 3.34, p = 0.029). Treating each symptom separately, but using Mann-Whitney U tests, there are significant differences for peak disorientation/ dizziness (U = 48, p < 0.001, one-tailed), peak nausea (U = 139, p < 0.047, one-tailed) and peak sweating (U = 94, p < 0.001), but not for peak general malaise (p > 0.1). As intended, vestibular stimulation produced motion sickness. However, the control procedure still scored equally on general malaise, demonstrating that it was appropriately aversive.

Four control subjects gave peak nausea ratings of 2.

Of the 31 subjects who underwent the treatment procedure, only 14 reported a peak nausea of 2 or more. These form the experimental group, with the other 17 forming the failed treatment group. Control subjects were yoked to the preceding experimental subject for time performing the treatment procedure (except that there was no control for S13). The results for the peak symptom scores for the experimental (Ex) and failed treatment (FT) groups are displayed in Table 3.17.
 

 


Mean (standard deviation)
None, not at all Slight Medium Considerable Extreme
Variable
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Ex
FT
Peak dis-orientation/ dizziness
2.9 (1.0)
2.4 (1.1)
0
1
1
2
4
5
4
6
5
3
Peak nausea
2.4 (0.6)
0.5 (0.5)
0
8
0
9
9
0
4
0
1
0
Peak sweating
2.1 (0.9)
1.5 (0.8)
0
2
3
5
7
9
3
1
1
0
Peak general malaise
2.1 (1.0)
0.8 (0.7)
0
6
4
8
6
3
2
0
2
0
Table 3.17: Peak symptom ratings in the vestibular stimulation group, split on peak nausea ratings.

Treating each symptom separately, using Mann-Whitney U tests, there is a significant differences for peak general malaise (U = 37, p < 0.001), but not for peak disorientation/ dizziness nor peak sweating (ps > 0.07).

Over all subjects, the four peak symptom measures were compared using Spearman’s correlations (Table 3.18).
 

variable

Peak nausea Peak sweating Peak general malaise
Peak disorientation/dizziness rS = 0.33;
p = 0.027
rS = 0.40;
p = 0.006
rS = 0.32;
p = 0.032
Peak nausea   rS = 0.56;
p < 0.001
rS = 0.49;
p = 0.001
Peak sweating   rS = 0.30;
p = 0.051
Table 3.18: Relationship between peak symptom ratings.

3.5.2.2.2 Responses to Kśrrta

The Experimental group consists of those subjects who underwent the treatment procedure and reported a peak nausea of 2 or more. 4 Control subjects (numbers 4, 9, 23 and 25) who reported peak nauseas of 2 have been excluded from the Control group; a further two (numbers 21 and 28) reported a peak nausea of 1. A third group, the Failed Treatment group, was not considered in the analyses.

A MANOVA was performed with a within-subject factor of the four kśrrta rating difference (of transformed data) scores (Contamination, How Hungry, Taste, Nausea) and a between-subjects factor of the two groups. This produced a significant interaction (F3, 19 = 4.20, p = 0.019). Each kśrrta question was separately compared between the control and experimental groups on t-tests: there was a significant result for the Nauseating difference score (t21 = 2.21, p = 0.039) and the How Hungry difference score (t21 = 2.35, p = 0.028); other tests not significant (ps > 0.1).

With all four kśrrta rejection measures being correlated, how can we be sure that we have a conditioned nausea and not just a secondary reaction to a hedonic shift in palatability? We can use the time 1 kśrrta scores to look at the relationship between the Taste and Nauseating scores: a regression of the time 1 Taste scores on the time 1 Nauseating scores was carried out, with all the subjects. This regression was significant (F1, 45 = 7.4, p = 0.009, adjusted r2 = 14%, n = 46), with the regression equation being:

Nauseating = -0.30(Taste) + 0.41

The small percentage of the variance explained is evidence in itself that the two are distinct.

Both parts of the equation are significant (ps < 0.01). This equation was then used to predict time 2 Nauseating scores from time 2 Taste scores, with the residual values (observed minus expected) then being noted. (These residual scores are equivalent to calculating a new Nauseating change score with the predicted time 2 Nauseating scores and then calculating a residual score of the difference between the two Nauseating change scores.) These residuals were then compared across groups: there was a significant difference (t21 = 2.52, p = 0.020).

However, perhaps we should exclude any subjects who were still feeling nauseous just prior to the second tasting. 3 subjects reported a Nausea rating of 1 then, one from each group (subject numbers 4, 18 and 32); only subject 32 was used in the above analyses. There is still a significant interaction on the MANOVA with the four kśrrta rejection measures (F3, 18 = 5.08, p = 0.010). On the separate t-tests, there were significant results with the t-tests for How Hungry difference score (t20 = 2.23, p = 0.037) and nearly for Nauseating difference score, t20 = 2.05, p = 0.054, but not for the others (ps > 0.1).

As very mild nausea may not elicit a non-zero response on the Nausea symptom question, we could also consider the General Malaise question. 1 additional subject reported a General Malaise score of 2 and 11 additional subjects reported General Malaise scores of 1 just prior to their time 2 tasting; only 9 scoring 1 are in the above analyses. This is, of course, a considerable reduction in n, leaving 7 Experimental, 13 Failed Treatment and 9 Controls only. There is still a significant interaction on the MANOVA: F3, 10 = 4.34, p = 0.033. However, there are no significant differences in the individual t-tests, ps > 0.1.

Differences scores were calculated between the two kśrrta ratings (time 2 minus time 1) and are shown in Table 3.19 for each experimental group. These figures exclude the Control subjects with peak nauseas of 2.
 
Mean
(standard deviation)
Median
Variable
Ex
FT
C
Ex
FT
C
Taste
-0.011 (0.179)
0.058 (0.237)
0.018 (0.081)
-0.033
0.033
0020
Contamination
0.043 (0.261)
0.074 (0.336)
-0.087 (0.216)
0.052
0.032
-.082
Nauseating
0.306 (0.397)
0.055 (0.208)
-0.013 (0.212)
0.214
0
0
How hungry
0.174 (0.287)
-0.066 (0.183)
-0.100 (0.246)
0.137
-0.054
-0.060
 

n
         
14
17
9
 
 
 
Table 3.19: Change in kśrrta ratings by group.

As opposed to comparing means between groups, we can analyse the results with correlations. For each group, peak nausea ratings were compared with the four kśrrta rejection difference scores (with the nauseous control subjects re-included). There were no significant relationships within the control group (ps > 0.1, ns = 13) or failed treatment group (ps > 0.08, ns = 17), but in the treatment group, peak nausea was significantly related to the How Hungry difference score (r = 0.68, p = 0.008, n = 14) only; other tests not significant (ps > 0.09). Including all subjects together, there is a significant correlation with the How Hungry difference score (r = 0.55, p < 0.001, ns = 44); other tests not significant (ps > 0.06).

Just using the treatment group, peak nausea was regressed on How Hungry difference score so as to create residual scores. This regression was significant (F1, 12 = 10.1, p = 0.008, adjusted r2 = 41.1%, n = 13), with the regression equation being:

DeltaHung = 0.30(PkNaus) - 0.56

where DeltaHung is the How Hungry difference score and PkNaus the peak nausea score. Both parts of the equation are separately significant (ps < 0.04).

This new residual score was now correlated with other measures (still just using the treatment group). There was no significant relationship with peak disorientation/dizziness or peak general malaise, but there was with peak sweating (rS = -0.59, p = 0.028, n = 14). There was no significant relationship with the past symptom measures (ps > 0.1). Nor was there any relationship with age, sex, trait STAI, Absorption, CNS, state STAI or hunger at times 1 or 2 (ps > 0.1).

Another way of measuring a conditioned nausea is to see whether nausea symptom ratings rose after the second consumption of the kśrrta in Experimental subjects. Change scores were calculated for symptom ratings at the beginning of the fourth physiological recording session minus those at the end of the third physiological recording session, that is, over the time in which subjects re-encounter kśrrta. Overall, there was little change in symptom ratings between these two times: in the experimental and control group (nauseous controls excluded), only three people showed change, all being in the experimental group and reporting an increase of nausea of one point. This is not significant on a Fisher exact test (p > 0.1).

Was there any conditioned reaction to the environs? This can be seen by considering the symptom ratings on returning for Session 2 versus the previous in between ratings: all changes were towards less nausea however.

3.5.2.2.3 Physiological Measures

Pulse:

The modal pulse frequency (on a spectral analysis) for each subject, for each of the four sets of physiological recordings, was recorded. Mean values are reported in Table 3.20.
 

Time
mean
standard deviation
minimum
maximum
n
Set 1 (baseline)
76
10
50
93
45
Set 2 (after ‘balance test’)
77
12
49
104
44
Set 3 (on returning for Session 2)
77
13
50
102
44
Set 4 (after re-tasting kśrrta)
77
16
50
112
44
Table 3.20: Pulse measurements (beats per minute).

Pulses from each set were very highly correlated with each other (0.88 > rs > 0.74, ps < 0.001, one-tailed, n = 43 or 44). We can use a reliability analysis to summarise the intercorrelations: for all four pulse scores, alpha = 0.94, n = 43.

A MANOVA was performed over all four times with a between-subjects factor of group (experimental, control and failed treatment): there were no significant results (ps > 0.1). The pulse was taken over 5 minutes, which may mask any transitory effect.

Electrogastrograms:

The electrogastrogram (EGG) recordings were analysed spectrographically. The reference ranges were 0-2.4 cpm (bradygastria), 2.4-3.6 cpm (normal) and 3.6-9.9 cpm (tachygastria), following Koch et al. (1994). Visual analysis showed many good recordings of the normal, cyclic pattern of stomach muscle activity, but other subjects’ recordings were poor.

Actual power ratings varied 104-fold: for normal frequencies, from 29µW to 271mW; for tachygastric frequencies, from 74µW to 475mW. The tachygastric power was divided by the power at normal frequencies to produce a ratio measure that should correlate with nausea. This varied considerably less (from 0.08 to 52), but the data still showed a large skew. Following visual analysis of these data and given that they are ratios, a log transform was performed. The data now appeared to follow a normal distribution.

The transformed data was analysed for outliers. Data from all four times was considered together and outliers excluded. The criterion for outliers chosen was that they were outside 1.5 times the interquartile range. Six outliers were so excluded.

It was thought that in more nauseous subjects, EGG recordings may have been less reliable, that is, it may not have been possible to obtain a clean signal for as long. However, there was no significant difference between the number of averages obtained from experimental versus control subjects at the second recording block (on a Mann-Whitney U test).
 

Recording block

Median

# averages

Mean... Mode...
1
3
2.4
3
2
2
2.0
2
3
2
2.4
3
4
3
2.6
3
Table 3.21: Number of averages available from the EGG recordings.

We can use a reliability analysis to summarise the intercorrelations. For all four EGG ratio scores, alpha = 0.31, n = 37. We would expect the least amount of change in Control subjects. Thus, the EGG ratio scores were correlated with each other for just Control subjects: this gave a perplexing pattern of correlations. EGG scores for recording blocks 1 and 4 showed a significant negative correlation (rS = -0.68, p = 0.029, n = 10), while the other correlations were non-significant (0.41 > rSs > -0.16; ns >= 11). On a reliability analysis, alpha = 0.48, n = 10.

The EGG scores for the second physiological recording session were correlated (using Spearman’s correlation) with the nausea symptom ratings given at the end of the treatment or control procedure. This result was not significant: r = 0.07, p >> 0.1. To adjust for interpersonal differences, a difference score was calculated between this EGG ratio score and the baseline figure. This new score also showed no significant correlation with the nausea measure: rS = -0.14, p >> 0.1. Splitting subjects by experimental group or excluding subjects with only one average taken for the EGG data for the second recording block made no difference to these results. Using a multiple regression analysis comparing the second recording block EGG scores to both the nausea symptom rating given at the end of the treatment or control procedure and at the end of that recording block also produced no significant result.

For the other recording blocks, associated nausea ratings should show less variance. Analyses as for the last paragraph for these other recording blocks’ produced similar non-significant results.

Log EGG scores over all four recording sessions were compared across experimental group (either experimental versus control or with failed treatment as well) on a MANOVA. There were no significant results by group or time.

Concentrating on the fourth recording block (after re-tasting the kśrrta), there was no significant correlation between the key kśrrta difference scores and EGG scores (on Spearman’s correlations).

It is always difficult to interpret negative results. As we would expect EGG scores to be related within each subject, the pattern of intercorrelations found suggests that there were problems in measuring EGG data. Also, any nausea produced may have been too transient to affect the EGGs sufficiently. I know of no other research that has attempted to measure such transient symptoms.

3.6 Conclusion

Although there are methodological problems in inducing LFAs, the second study (3.5.2) demonstrated that a LFA induced to a food resulted in that food becoming rated more nauseating. There was no signficant change in the palatability question. With a small study like this, it is not safe to conclude that there is no hedonic shift in palatability, but there clearly is a learned nausea of some sort. This concords with anecdotal evidence.

This result must call into question the utility of LFAs as a model for palatability change, but, conversely, it furthers suggestions of links with ANV and other learned nausea phenomena. Future research will have to explore these potential links more explicitly. Many nauseas of no apparent organic origin could be psychological in nature and the demonstration of how easy it is to produce a learned nausea here is surely relevant.

It is apparent how a learned nausea could act as reinforcement for a LFA otherwise challenged in extinction. This perhaps explains the durability of these responses. While individual LFAs are rarely a subject for medical concern, their role in ‘cancer anorexia’ means that it would be useful to be able to extinguish them. Could anti-nausea drugs be of use? Again, further research is needed in devising schemes to help reverse these aversions in anorectic patients.

It is only by considering the nature of the aversion that these matters arise. The ubiquitous use of rats and mice in this research can be seen as an unfortunate accident. These species cannot vomit, removing our main way of indexing nausea in animals. New animal models for nausea and vomiting have been developed and will hopefully come to replace rats and mice where appropriate.