Measuring hedonic impact in animals and infants: microstructure of affective taste reactivity patterns

Abstract

The hedonic impact of taste is reflected in affective facial reactions made by human infants, other primates, and even rats. Originally studied in human infants, affective reactions to taste have also been used by affective neuroscience to identify hedonic brain systems in studies of animals (via application of neural stimulation, pharmacological activation, and neural lesion manipulations). The chief limitation of measuring affective reactions is that it is difficult for experimenters to know how to interpret them, and therefore how to interpret changes produced by brain manipulations. This paper notes guidelines to interpretation. It examines the phylogenetic continuity between humans, other primates, and rats in terms of the microstructure of taste-elicited affective reactions. It reviews evidence that affective taste reactivity patterns truly reflect a ‘core hedonic process’ of palatability or affect, rather than being an ingestion measure, consummatory behavior measure, or a sensory reflex measure. It reviews affective neuroscience studies of taste reactivity that have identified true hedonic brain substrates, and discriminated them from false hedonic brain substrates. It considers the neural bases of incentive ‘wanting’ versus ‘liking’. Finally, it notes the difference between human subjective affective ratings of pleasure and ‘core hedonic processes’ reflected by behavioral affective reactions.

Introduction

We can learn much about both humans and animals from the microstructure of their behavioral affective reactions, as Darwin pointed out a century ago [1]. This paper concerns the nature and measurement of taste-elicited affective reactions, and their use to study affective brain mechanisms. I will focus on taste reactivity as a measure of hedonic impact in humans and animals. For discussion of other aspects of taste reactivity microstructure (related to intake, ingestion pattern, or homeostasis), see studies and reviews by Grill, Spector, Kaplan, Flynn, Norgren, and their colleagues [2], [3], [4], [5], [6], [7], [8].

The first of the modern wave of taste reactivity studies appeared around 1970, when Jacob Steiner published photographs of the facial reactions of newborn human infants to sweet, salty, sour, and bitter tastes [9], [10]. Taste reactivity measures have subsequently been applied to many topics in the psychology, physiology, and neurobiology of motivation and regulation [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35]. In the original studies, Steiner strove to test his infant subjects typically within a few hours of birth, before their first postnatal feeding. He was concerned to demonstrate what the human newborn was capable of ‘instinctively’, in advance of postnatal learning about foods and their nutritional consequences. Thus the infants had absolutely no experience with tastes or their consequences (aside from prenatal swallowing of amniotic fluid) before the first squirt of sucrose, quinine, or another solution into their mouths.

Infant facial reaction patterns were originally considered to be sensory-typical by Steiner, who described them in terms of the eliciting sucrose, salt, quinine, and other tastes that triggered particular patterns [9], [10], [33]. But there were essentially just two types of pattern. Sweet sucrose elicited positive or hedonic¹ patterns of lip smacking and tongue protrusion, accompanied by relaxation of the muscles of the middle face, and an occasional smile. Bitter quinine elicited negative or aversive gapes, and complex grimaces involving retraction of the lips, ‘scrinching’ of the brows and muscles around the eyes, and ‘wrinkling’ of the nose (also flailing of the hands and arms, and small shaking or retraction movements of the head away from the taste). Salt, sour, and other tastes evoked various degrees of intermediate reactions between these extremes.² Steiner's original observations on human infants have since been replicated several times, both by him and his colleagues and by other groups [19], [29], [31], [35]. Subsequent studies have improved the data resolution through video-recording and fine-grained analyses, but the basic phenomenon has remained the same (Fig. 1). And in the first affective neuroscience application, Steiner went on to show that sweet and bitter tastes elicited the same positive and negative facial reactions from anencephalic infants and hydrocephalic infants, who were born with congenital malformation of the forebrain (or absence, in the case of anencephalics) [9].

Early studies emphasized the sensory aspects of taste reactivity patterns in both human infants and animals [9], [22]. Only in the following decades did it become clear that facial affective reactions to taste reflected the hedonic valence rather than the sensory identity of the taste, when physiological, psychological, and pharmacological manipulation studies of animals separated sensory versus affective aspects of taste stimuli. Sweet and bitter usually trigger polar opposite patterns of affective reactions. Sour, salt, water, and other tastes typically evoke less intense mixtures of the two polar patterns, rather than distinct reactions corresponding to their own sensory quality. There are no unique identifying reaction patterns for sweet, salt, sour, bitter, umami (a special protein taste linked especially to glutamate that may exist for both humans and other animals [36], [37], [38], [39]), or other basic tastes. If taste-specific patterns existed, that would imply sensory coding of reactions, and would allow an observer to infer the sensory content of the taste reliably from watching the reaction. But with the possible exception of lip pursing (which may be a distinct human reaction to sour [29]), no facial reaction is absolutely specific to a particular taste sensation. “Bitter-typical” expressions can be elicited by concentrated salty, sour, and other tastes. “Sweet-typical” reaction patterns can be elicited by milk or other complex tastes, by dilute NaCl, and by other tastes. An observer therefore cannot tell the precise sensory quality of a taste based on an infant's reaction. By contrast, however, an observer can make an excellent inference concerning whether an human infant ‘likes’ a taste based on her or his facial expression—that is, infant facial reactions seem to be an excellent indicator of taste palatability or hedonic impact (Fig. 1) [35].

Human infants are not the only members of our species, of course, to display distinctive facial expressions to positive or negative tastes. Adults do so too. Human infants, however, may be more responsive and ‘honest’ in their facial expressions than older children or adults. That is perhaps because of the socialization of human emotional expression in early life, and because human voluntary control over facial expression can prevent human facial expression from being an accurate indicator of underlying hedonic impact [30]. As social individuals, we may suppress real emotional reactions or counterfeit false ones. Thus studies of adults or older children have been less successful than infant studies in using facial expressions to measure affective reaction to tastes or odors [20], [32], even though adult facial expressions do often express the hedonic impact of food [40]. The facial reactions of some human adults are less constrained by social motives than is ordinarily the case, and Steiner and his colleagues have obtained clear hedonic versus aversive patterns in Alzheimer's patients and in certain other neurological populations [41]. Yet as a general rule, the facial reaction measure of hedonic impact may be most sensitive when applied to human infants or to animals from other species (of any age).

In a series of comparative studies, Steiner and Glaser examined taste reactivity microstructure in great ape, old world monkey, new world monkey and prosimian infants and adults [34], [35], [42]. Virtually all of those primates emit facial reactions that are appropriate to the human-judged palatability of bitter, sweet, or other tastes, and similar to reactions of human infants. For example, quinine typically evoked gapes and other aversive responses from chimpanzees, gorillas, orangutans, and monkeys [34], [42]. By contrast, sucrose evoked tongue protrusions, mouthing movements, and sucking, and other tastes evoked appropriate intermediate responses, from most species of primates [34], [42].

The comparative microstructure of human infant and other primate taste reactivity was analyzed explicitly in a recent comparative study by Steiner et al. [35]. In that study, affective reactions to tastes were compared among 12 different primate species (including humans, great apes, Old World monkeys, and New World monkeys). If one believed that the facial expression of hedonic impact is qualitatively unique to human beings then one might have expected that human infant facial reactions would be quite different from all other primates, and that other primate species might be less differentiated from each other. However, that was not the case. Instead, human infant facial reactions to sweet, bitter, and other tastes were found to be closely similar to those of chimpanzee, orangutan, and gorilla. When components were plotted to reveal similarity among species, a single ‘hominoid cluster’ was formed that contained both human and great apes (humans and great apes together belong to a single phylogenetic group called hominoids, whereas all monkeys belong to different groups). In terms of microcomponent structure, human taste reactions were more similar to great apes, than great apes were to monkeys (Fig. 2). Old World monkeys that evolved in Africa or Asia formed a separate cluster of their own, while New World monkey species that evolved in South America formed a third cluster (this cluster also seemed closest to the single prosimian species that was tested).

The results of Steiner et al.’s [35] comparative analysis indicates that some components of taste-elicited reactions are universal among all primates, such as gapes to bitter or rhythmic tongue protrusions to sweet. Other components belong only to particular species or groups of species. For example, only humans and other hominoids such as great apes responded to sweet tastes with complex lip smacking, or responded to bitter tastes with the complex ‘scrinching’ movements involving musculature of their brows, nose, and middle face. Conversely, only South American monkeys appear to show affectively coded shapes of tongue protrusion (forming an upward pointing trajectory to sweet but a downward trajectory to bitter) [35].

The upshot of all this is that the particular configuration of components shown by a particular primate is largely a function of genotype. No two primate species are identical in the microstructure of their facial reactions to taste, but the degree of difference between them is continuous rather than categorical, and is proportional to the phylogenetic distance between the species.

The evolutionary continuum of taste reactivity microstructure extends to nonprimate animals as well. Grill and Norgren published a landmark study in 1978 of the behavioral taste reactivity components emitted by rats [22]. The affective reactions of rats are distinctly related to those of primates (Fig. 2). Infusions directly into the mouth (through an implanted oral cannula) of bitter quinine elicited gapes, head shakes, forelimb flails, and other reactions from rats, which Grill and Norgren termed aversive, following Craig's 1918 definition of aversion as “a state of agitation which continues so long as a certain stimulus, referred to as the disturbing stimulus, is present” [43; p. 91]. These reactions were scored in fine detail by Grill and Norgren in a frame-by-frame video-analysis. Conversely, sucrose elicited a different pattern of reactions Grill and Norgren called ‘ingestive’, including rhythmic tongue protrusions and mouth movements, and ‘lateral tongue protrusions’ (somewhat like a lateral licking of the lips or chops) [22], [23], [44]. Hamsters also show similar positive reactions to sucrose (lateral and midline tongue protrusions) and aversive reactions to quinine (gapes, chin rubs, headshakes, forelimb flailing, etc.) [45].³ Subsequent studies of taste reactivity patterns have added paw licking to the list of positive or ‘ingestive’ sucrose-elicited reactions for rodents [21], [46], compiled procedures for comparing humans and nonhumans [35], and developed methodological improvements, such as time-bin scoring procedures to balance hedonic and aversive categories of reactions [21], [47], [48], and methods to avoid the masking of palatability shifts by response demand properties of the taste [2].

Aversive gapes and positive hedonic tongue protrusions are universal affective expressions, emitted by human infants, other primates, and rats. To say that the ‘same’ microcomponent (for example, rhythmic tongue protrusions to sucrose) is emitted by two different species is not to say that the movements are identical. After all, physical anatomy of the ‘same’ part varies across different species. For example, all mammals have a head, but its shape and size may be different in different species. It is not surprising, then, that behavioral morphology varies systematically too, even for the ‘same’ (i.e. analogous and homologous) taste reactivity microcomponent. The important point is that the variation must be merely systematic if the microcomponent is to be considered truly the ‘same’.

An instance of ‘merely systematic variation’ is the scaling of component speed to body size. The duration of the ‘same’ stereotyped reaction in different species appears to be directly proportional to the average adult body size (mass) for that species. Such size-based timing rules are called allometric, and are very common in physiology and behavior (the duration of heartbeat rhythms, walking step cycles, etc. [49]).

Certain taste reactivity components also follow an allometric timing rule, especially those that are highly rhythmic and/or stereotyped in duration. A good example is rhythmic tongue protrusion to sucrose, which by definition follows a precise rhythm. Another example is gape to quinine, which is quite stereotyped in duration. Allometric timing can give rise to a perceptual illusion of difference between species of different sizes. For example, when one compares rhythmic tongue protrusions made by a human infant with rhythmic tongue protrusions made by a New World Saguinus or Callithrix monkey, the rhythm is so different in speed as to appear almost to be a totally different reaction. Human infant tongue protrusions are slow and almost languid, about 1 per second, whereas the South American monkeys emit rapid-fire tongue protrusions more than three times faster. Yet both species follow the same generative timing rule in which movement duration is directly proportional to body mass [35]. These rates correspond to an average human adult body mass of around 80 kg, whereas the average adult body mass for these New World monkeys is only about half a kilogram.

The allometric timing rule can be stated in the form of an equation. For primates, and perhaps all mammals larger than about one-half kilogram, the timing of rhythmic tongue protrusions is: duration (sec)=a×(average body mass in kg)^b, where a and b are both constants $\text{(a=0.27,}\text{b=0.32)}$ [35]. Steiner et al. found that the actual duration of primate tongue protrusions closely follows this allometric rule (r=0.88; Fig. 2). Therefore gorillas (weighing about 100 kg) have tongue protrusion cycles even slower than humans (approximately 1000 ms), whereas chimpanzees (weighing about 60 kg) are slightly faster (approximately 850 ms) than humans [35]. Old World monkeys are faster still (approximately 480 ms on average), and tiny New World monkeys have the fastest tongues of all primates (approximately 200 ms) [35]. In general, as average adult weight increases across species, the b exponent value of 0.32 means that the duration of cycles grows roughly in proportion to the increase in the cube root of body mass. A similar allometric rule applies to the duration of gapes [35].

Although the duration of the components are predicted on the basis of adult body mass (Table 1), the rule also appears to apply equally to infants of each primate species [35]. Infants have components that are similar in duration to adults of the same species, despite the size difference between infants and adults. Thus human infants follow the adult rule in their timing (corresponding to adult body mass, and not to their own small body mass), and so do infant New World monkeys [35]. This suggests that the timing evolved parameters evolved to coordinate adult movements, perhaps because they have the greatest impact on fitness. It means also that the timing parameters are programmed into the brain for both infants and adults. The duration cannot result passively from the physics of movement, since if it did small infants would be faster than adults. The only alternative is that timing must be actively imposed by neural central pattern generators. Infant brain timing circuits are genetically coded in advance to conform to the species-typical size they will have if they grow to be adults. Similar infant anticipation of adult timing, and genetic programming of timing parameters, has been found for other types of stereotyped movement [50].

Interestingly, it appears allometric timing rules for hedonic and aversive reactions might also extend beyond primates, connecting humans to rodents. For example, it is interesting to note that Grill and Norgren [22] originally reported that rats have a timing of approximately 120 ms for the duration rhythmic tongue protrusion cycles (in our lab, we generally obtain a range of 120–150 ms for rats that weigh approximately 300–400 g; the equation predicts 180 ms). And mice, weighing only about 30 g, have a tongue protrusion lick cycle of about 90–110 ms duration (mouse lick cycle data provided by K.N. Hewitt and Dr Peter Clifton, personal communication, August 1999 and Ref. [51]; the equation predicts 85 ms).

Of course it would be necessary to do an explicit comparative study before drawing a strong conclusion, but these values correspond rather well to the extrapolation from the primate allometric line shown in Fig. 3. I have repeated the nonlinear regression analysis of Steiner et al. after adding rodent data to the primate data, and find little change in the values for the allometric equation (y=a×x^b). The exponential b value remains 0.32 in both cases, and the value of the a constant changes from 0.27 to 0.26. This is hardly any change at all, and the correlation between duration and body weight (r² value) actually improves from 0.88 to 0.91 after adding the rodents to the primates. Thus the underlying relationship between timing and body mass appears to span from gorillas to mice. Even though a gorilla weighs 3000 times more than a mouse, each follows an equivalent allometric timing rule, the parameters of which are likely programmed into the motor circuitry of their brains.