Minutes of the Columbia University Seminar on Appetitive Behavior (#529)

Chairman: Harry R. Kissileff, Ph.D.
Rapporteur: Janet L. Guss
Date: Thursday, March 9th, 2000
Title: “Pavlovian Occasion Setting: A Link Between Physiological Change and Appetitive Behavior”
Speaker: Terry L. Davidson, Ph.D., Department of Psychological Sciences, Purdue University

Presentation (provided by speaker)

Researchers have long attempted to identify physiological mechanisms that underlie the initiation and termination of eating and behavior instrumental to obtaining food. However, less effort has been devoted to specifying the links between physiological activity and behavioral change. In fact, the mechanisms that are responsible for adaptive behavioral output in response to changes in the status of animal’s energy resources are often represented by nothing more than an arrow, a “+” or a “-“ in a diagram. The purpose of this communication is to propose that the links between physiological activity and control of feeding behavior may be understood in terms of contemporary Pavlovian learning and memory principles.


Simple excitatory conditioning.

Memory refers to the activities of the brain that are involved with encoding, representing and organizing experiences. Pavlovian conditioning provides a basic methodological and conceptual framework for describing how memorial representations of such experiences are retrieved or activated (see Holland, 1990). Within this framework, a conditioned stimulus (CS) is established as a signal for the occurrence of an unconditioned stimulus (US). As a consequence of this training, the CS becomes associated or linked with the representation of the US in memory. Evocation of conditioned responses (CRs) by the CS is usually viewed as evidence that the presentation of the CS has excited or activated the memorial representation of the US. Because they occur before or “anticipate” the delivery of the US, CRs are sometimes described as anticipatory responses.

Although Pavlovian conditioning was first demonstrated with punctate CSs (e.g., a brief auditory cue) that preceded, with little or no delay, the presentation of discrete USs (e.g., a small amount of food), the same principles apply to learning about diffuse, long duration CSs and USs, that are separated by long temporal delays. For example, learning to avoid toxins and to seek-out and consume nutritious foods rather than nonnutritious substances can be easily described within the Pavlovian framework. That is, orosensory stimulation (e.g., taste) can be viewed as a CS that precedes, in some cases by several hours, the experience of postingestive sensory stimulation. Based on this relationship, the orosensory CS becomes directly associated with the memory of a postingestive US. If the postingestive US is aversive (e.g., intragastric malaise) the orosensory CS evokes avoidance or defensive CRs. If the postingestive US is appetitive (e.g., involves the restoration of nutrient or energy balance), the CS evokes approach and consummatory behaviors.


Pavlovian occasion setting.

Pavlovian occasion setting involves a learning process distinct from the simple conditioning process described thus far. An occasion setter is not itself a signal for the occurrence of a US, but instead signals that a CS will be followed by US presentation. Occasion setting is demonstrated when animals exhibit more CRs (e.g., approach a food cup where the US is delivered) on trials in which the occasion setter precedes the presentation of the CS compared to trials in which the CS is presented without the occasion setter. Unlike excitatory CSs, occasion setters evoke little or no conditioned responding when they are presented alone. Thus, ability of an occasion setter to promote responding does not appear to depend on its own direct excitatory association with a US or a response. Moreover, the response promoting power of occasion setters is limited largely to target CSs that have been trained both with and without reinforcement. These training conditions serve to embed the CS in concurrent excitatory and inhibitory associations with the US. The inhibitory association opposes the tendency of the excitatory association to activate the memory of the US, thereby making it more difficult for the CS to evoke conditioned responses. Occasion setters are thought to promote conditioned responding by removing or reducing the strength of this inhibitory CS-US association (see Swartzentruber, 1995).


Occasion setting and the control of feeding behavior.

The regulation of feeding behavior can also be described within an occasion setting framework. Internal changes related to departures from energy balance are assumed give rise to discriminable sensory consequences. These “hunger stimuli” were once described by Cannon as a “very disagreeable ache or pang or sense of gnawing or pressure which is referred to the epigastrium.” Although recent discussions do not identify the sensory aspects of hunger with any specific physiological locus, these signals could be the result of changes in the availability or utilization of metabolites, changes in neuropeptide activity, or shifts in the type or source of energy (gut, adipose tissue) utilized, among other possibilities. Regardless of their origins, hunger stimuli may function as occasion setters by signaling that food-related CSs will be followed by a particular pattern of postingestive stimulation. For example, when hunger stimuli are present, orosensory CSs (e.g., tastes) produced by contact with food are followed by an appetitive postingestive US (i.e., a US that is capable of promoting appetitive behavior) that is assumed to be linked to the restoration of energy and perhaps macronutrient balance. When hunger stimuli are absent, those same orosensory cues are no longer followed by that appetitive US. Thus, just as a conventional Pavlovian occasion setter is informative about the reinforcement of its target CS, hunger cues are also informative about the postingestive reinforcement of orosensory and other food-related cues.

Consistent with the general occasion setting analysis, hunger stimuli are assumed to potentiate the performance of conditioned responses by reducing the strength of inhibitory associations between food-related CSs and the appetitive postingestive US. Functionally, this means that hunger can potentiate feeding behavior only when the food CSs that are present have both excitatory and inhibitory associations with the memorial representation of the postingestive US. How are these concurrent excitatory and inhibitory associations formed? The excitatory association between food cues and the appetitive postingestive US is formed when the animal eats food when it is hungry. The inhibitory association is formed when animals eat food in the absence of hunger and thus do not experience that appetitive postingestive US. Examples of eating when hunger is absent may be based on learning, hedonic, or other nonhomeostatic processes. When eating occurs in the absence of hunger, food CSs activate the memory of the postingestive US, but the sensory postingestive US that is normally present when the animal is hungry does not occur. That inhibition would develop following such expectancy violations is consistent with several theories of animal learning. Hunger stimuli promote appetitive behavior by removing or weakening the inhibitory food CS-postingestive US association.


Empirical Support.

Several types of evidence are consistent with this analysis. These include: (1)Different durations of food deprivation give rise to distinct interoceptive stimuli . Rats must discriminate the interoceptive sensory consequences of different durations of food deprivation because they can solve discrimination problems when cues arising from different durations of food deprivation are the only relevant stimuli. Moreover, discriminative control of responding based on these food deprivation cues generalizes to interoceptive stimuli produced by physiological manipulations (e.g., administration of antimetabolites) that are known to influence food intake (see Davidson, 1993). Therefore, stimulus control is based on internal cues, rather than on external cues related to the deprivation regimen.

(2) Hunger cues modulate the capacity of Pavlovian CSs to activate US representations.There are two ways in which occasion setters could affect responding: summation or modulation. According to the summation explanation, the CS and the occasion setter are partners in association with the US. If either is presented alone the animal will not respond, but if both are presented together the animal will respond. If the animal is then given nonreinforced trials (i.e., extinction) with the CS alone the occasion setting stimulus would sum with a CS whose effectiveness is reduced and together they would produce less responding. In contrast, according to the modulation explanation, prior nonreinforced trials with the CS alone would not necessarily reduce responding when the occasion setter and CS are presented together. That is because the occasion setting stimulus promotes responding by weakening the inhibition that would develop to the CS as a result of extinction trials. The available results support the modulation account. For example, occasion setters have been shown to promote more responding to a CS that had been given reinforced training, followed by extinction trials (i.e., a procedure that would produce inhibition) than to a CS that had not received extinction trials. These results have also been obtained in studies that trained rats to use stimuli arising from 24-hr food deprivation as occasion setting cues (see Davidson, 1998 for a description of these and other related findings).

(3) Occasion setting by hunger cues depends on the existence of inhibitory CS-US associations. According to the modulation explanation, animals learn inhibitory associations between food CSs and postingestive USs as a consequence of eating food when satiated. If hunger cues promote appetitive behavior by reducing the strength of these inhibitory associations, appetitive behavior evoked by food CSs that have not been given this inhibitory training should be quite insensitive to either increases or decreases in hunger. Results from our laboratory and elsewhere are consistent with this prediction. For example, animals must first eat a specific food when satiated before satiation can exert a strong suppressive effect on appetitive behavior evoked by a CS for that food (see Davidson, 1998). This interpretation is also supported by findings that the suppressive effects of food satiation on appetitive behavior (but not intake) are weaker for rats with hippocampal lesions than for controls. Hippocampal lesions also appear to retard learning of simple inhibitory associations in rats.

Conclusions. Pavlovian occasion setting may provide a conceptual link between the physiological and the behavioral controls of feeding behavior. Within this account, hunger stimuli modulate the effectiveness of associations between excitatory food CSs and postingestive unconditioned stimuli (USs) by reducing the strength of concurrent inhibitory associations that also develop between those cues. This approach accounts for many effects of hunger on feeding behavior without appealing to motivational concepts. Instead, the principles used to explain regulatory feeding behavior are the same as those that describe how animal solve conventional discrimination problems.

References


Discussion:

General Question: What is the mechanism by which hunger promotes the performance of appetitive and consummatory behaviors?
Does hunger promote feeding behavior by enhancing the reward value of food?
A: “Reward value” is just a descriptor or ‘label’. How does hunger cause an animal to eat more food (or rate food as more pleasant, or prefer it, or make responses to obtain and ingest it)? By enhancing the value (or palatabililty, or hedonics) of food. How do we know that the value of food increased? Because the animal eats more of it (or rates it more highly, or shows more appetitive taste reactivity responses to it, and so on). This is circular. Although manipulations that are presumed to produce hunger have been linked causally to modifications in the strength of appetitive and consummatory behavior, what independent evidence links changes in incentive value to changes in hunger?

Geary: If you believe that terms such as ‘reward value’ are circular and don’t provide mechanistic explanations of behavior, then do you believe that the behavioral analyses used to study associative learning and operant conditioning are also ‘circular’?

A: Yes- Reinforcement can be as circular as the terms “reward value” and “motivation”. An explanation is circular when the only evidence for the existence of a proposed explanatory process or mechanism is the occurrence of the phenomenon that the process or mechanism is supposed to explain. The term reinforcement, like reward value, is often used without empirical referents that are independent from the phenomena to be accounted for. I try to avoid these types of terms in my model. An advantage of using a Pavlovian model is that the conditions that result in association formation and the behaviors that serve as the index of learning are relatively well-specified empirically. Applying principles that have been useful to predict and explain phenomena in one domain, to predict and account for changes in feeding behavior involves more than labeling. I am trying to add more information to the model, such that we can explain behavior rather than just apply labels.

Schwartz: Are you using hunger in this model as the occasion setting model of food intake, or is this the occasion setting model of hunger? Also, please explain what you mean by hunger.

A: By hunger I am referring to a stimulus or set of stimuli that is produced by departures from energy homeostasis that result in negative energy balance. These departures can and usually do occur as a normal consequence of a period without food intake or as a result of reduced availability of food. My idea is that what animals learn about their interoceptive hunger stimuli enables them to predict accurately the postingestive consequences of eating. That is hunger stimuli “set the occasion” for the occurrence of certain types of postingestive stimulation.

Schwartz: But there are many situations incurred by animals or humans in which they are motivated to eat, yet we have no reason to believe there has been any change in their ‘energy state balance’....

A: Yes- that is why I am only speaking of situations that are independent of non-regulatory controls, or scenarios.... So, for example, learning is used to explain non-regulatory controls in many situations without reference to hunger. I’m trying to set up a model that can explain how the regulatory control of eating might also involve learning mechanisms. The mechanism underlying the regulatory control, but not necessarily the nonregulatory control, of eating is presumed to involve learning about hunger cues.

Kissileff: So you’re saying hunger is a label for an operation that you’re attempting to use in this model and the particulars by which hunger is generated are not the subject of your inquiry?

A: Yes, for the purpose of this model the physiological mechanism responsible for the state of hunger is not germane. I assume that hunger arises as a result of physiological change and I have done studies aimed at identifying the events that give rise to hunger signals. But within the present model, I am interested in specifying how the sensory consequences of those physiological changes, whatever their origins, influence behavior.

Collier: In your model you suggest the animal has to learn the connection between the CS and the US, but isn’t it the case that the CS is associated with the UR? Why is your paradigm different?

A: That is the stimulus-response (S-R) interpretation of Pavlovian conditioning that was proposed by Hull and other early behaviorist. However, I think the stimulus-stimulus (S-S) view that was promoted early on by Tolman and others is favored by most contemporary Pavlovian theorists. One reason the S-R interpretation has become less popular is that the conditioned response often bears no resemblance or even opposes the UR. Although not ruling out the possibility of S-R learning, these findings make it difficult to see how the CS could be directly associated with the UR.

Geliebter: With regard to the inhibitors, are they similar to extinctions?

A: Extinction is a procedure that produces inhibition. The inhibitor makes it harder for an excitatory CS to activate the memory of a US. That is, inhibition opposes excitation thereby making it harder to reach the excitatory threshold for activation of a memory event.

Mook: Are you suggesting that extinction is just a matter of forgetting the original association?

A: No, that is what Rescorla said as part of the Rescorla-Wagner model and that just doesn’t work, because when you extinguish the association in an animal and then give him one trial, he is right back where he was after he’d been trained. If extinction were simply forgetting, we’d have to start him up again. Therefore this ease of reactivation of the memory argues against Rescorla’s notion.

Mook: Well if you make this a matter of memory activation, that is, if that’s what inhibitable, then doesn’t that objection, or case against the model, also apply to the model you are proposing?

A: I don’t think so, because if you have that inhibition intact, that inhibition can be removed; that’s what I am saying the occasion setter does. The occasion setter reduces the learned inhibition and lowers the threshold for memory activation. In other words, by removing or weakening inhibition, an occasion setter makes it easier for an excitatory CS to activate the memory of the US.

Schwartz: The word ‘memory’ in your model seems as circular as the word ‘value’ or ‘motivation’, and actually analogous to the suggestion you made earlier; i.e. that these words were circular.

A: Possibly the distinction will become more clear as I elaborate on my model.

Sclafani: How much of your theory depends upon the UR?

A: The theory depends on the UR to the extent that some change in physiological activity occurs.

Sclafani: But there’s no evidence here that any physiological change has occurred in your model.

A: This is an interesting thought because it is also the premise of your nutrient conditioning paradigm. For the purpose of my model, the particulars of the UR are not important. Within most contemporary Pavlovian accounts, the predictive relationship between the CS and US produces an association. The evocation of CRs provides the critical evidence that the CS-US association has been formed. This relationship does not depend on the nature of the UR or on whether or not the UR is even specified. So in this model the UR is irrelevant.

Geary: Do the food cues activate the memory of the postingestive consequences? Because postingestive consequences take a long time to develop.

A: There are many different kinds of postingestive consequences, and once you change the internal state of the animal, any of those signals can be viewed as postingestive consequences.

Kissileff: Can you provide us with an example of a postingestive US?

A: One can assume that a postingestive US is any change in state relative to the pre-fed state, that is post-orosensory. I’m not concerned with the particulars of those postingestive stimuli, but I would imagine we are all experiencing them now, since we just ate a meal. I don’t think anyone has determined yet what events constitute the postingestive US.

Kissileff: Why is the post-ingestive event labeled a US instead of a UR?

A: This is a good question: The postingestive stimulation (a US) is probably produced by a bodily response (i.e. a UR). However, the general occasion setting model focuses on what animals learn and remember about the stimuli they experience. Within this context, the important thing that an animal would learn about a UR relates to the stimuli that the UR produces. That is probably why the focus is on the stimulus (US) rather than the response (UR) properties of the postingestive event.

Geliebter: With regard to the light-intensity experiments, were the animals trained originally to associate the consequence with a high intensity light?

A: We had two groups, and half were trained with high intensity light first while the other half received low intensity light first.

Houpt: Can you provide more details about the intensity of the light stimulus?

A: Actually rats are more sensitive to auditory stimuli so the high and low intensity stimuli used in this study were auditory, which then was followed with the light cue, which then resulted in access to food.

Kissileff: What was the response measure in this experiment and was there another condition?

A: The response measure was whether or not they approached a food cup in response to the occasion setting stimulus (the sound followed by the light).

Kissileff: Did they ever get food from the cup?

A: They were trained with the food cup, and food was the US in this model, but whether or not they actually took food and ate it from the cup was not a focus of the study. I do know that the food cup was always empty at the end of each training session.

Schwartz: So in this ‘light-sound experiment’, you’re using the tone intensity as the occasion setter; but how would you characterize what you’re using as hunger? Can hunger be used as a CS?

A: I suppose it could.

Sclafani: It should be noted that not everyone agrees that food is simply a reinforcer with a reward value.

*Speaker’s Comment (Davidson): In my model I am simply showing that hunger acts as a memory activator, to help the animal recall the association between the CS (food) and the US (postingestive state). This model is not one based on reward.

Guss: Can you explain why you don’t consider the hunger state of the animal a motivational state that ‘motivates’ the animal to remember the association?

A: I would not call it that because I don’t know if hunger is in fact motivational; all I can say is that hunger is a stimulus.

Mook: You keep referring to what the consequences of feeding are; why does the animal ‘care’ about those consequences? In other words, if you aren’t going to use a motivational mechanism to account for these data, how do you separate the association that the animal is disposed to act upon from what it is relative to?

A: Why does the animal have to care about the relationship at all?

Mook: Because if the animal experiences the postingestive consequence after receiving the food when it is hungry, one would assume the animal makes the association: “if I eat, I experience these consequences, which remove the hunger, so I think I’ll eat...” instead of saying “Isn’t that interesting?”

A: The way I’ve defined this model is such that I have two classes of behaviors: appetitive and aversive. So I can produce appetitive behaviors by putting meat powder in the animals mouth and aversive behaviors by putting quinine in its mouth. Therefore different stimuli elicit different classes of behavior. By defining some of these behaviors as ‘motivational’ just doesn’t tell me much more than I already know. That is the basis of my strategy.

Kral: I don’t think you’re making a very convincing argument that there’s a difference between motivation and memory in this model.

A: I think memory and motivation are one in the same, and I agree with you 100% in the sense that they both do the same thing. The reason I prefer to use the memory model is because I can be sure that the animal remembers the associations that were learned--that is there is much independent evidence for the operation of learning and memory processes-- but I can’t say for sure that the animal is motivated by those memories. What is the evidence for motivation that is independent of the phenomena that motivation is supposed to explain? Reliance on the term motivation seems to be an over interpretation that is not required by the data.

Sclafani: In a way you are implying that changes in the value of food does occur; the postingestive consequence has a different action when the animal is in the hunger state, so there’s a change in the value.

A: No, all I’m saying is that the eating produces a particular US depending on the state of the animal. I don’t think you have to assume the event (i.e. getting food) is more valuable when the animal is hungry, it’s just easier to remember the US. One need not assign value to that event. The postingestive US that the animal receives when it is hungry differs from that which it receives when it is not hungry.

Schwartz: No one is saying the value has to be ‘better’ or ‘worse’ but any qualitative change in the memory of this postingestive US is a value change. Value also implies some qualitative change in something- not necessarily positive or negative, but why can’t the word value be used?

A: Because hunger does not necessarily change qualitatively the memory of the US-I’m not suggesting that. No capacity of hunger to change the value of anything is part of this model. Hunger is simply a retrieval cue that determines which types of memories are likely to be recalled. In addition, the mechanisms involved with the retrieval of memories and the conditions necessary for memory retrieval (e.g., the existence concurrent excitatory and inhibitory associations seems to be more clearly specified within the occasion setting framework, than are the mechanisms that underlie changes in reward value in other models.

Collier: Memory is not something you can measure in an animal; it either approaches the food or it doesn’t approach the food. Therefore your use of the word memory seems redundant. If you don’t want to commit to any description of the postingestive event because you can’t measure that, and you can’t measure ‘value’, how can you say you are measuring memory activation?

A: I think that the usefulness of memory, learning, motivation, incentive value-any hypothetical construct--can only be assessed with reference whether the construct is used in a way that leads to interesting testable, predictions and to novel interpretations of previously observed phenomena. The occasion setting model has provided a novel, testable account of performance in conventional Pavlovian discrimination problems. We say memory activation is involved in these problems because this conceptualization currently provides the best available account of the data. What I am proposing is that the regulatory control of feeding also depends on the ability of animals to solve a similar type of discrimination problem. This problem is solved when animals learn that their interoceptive hunger stimuli signal that food CSs (e.g., tastes or other food-related cues) will be followed by a particular type of postingestive US. Even though no one can specify as of yet, the precise physiological locus or nature of the hunger cues and postingestive USs that are part of this model, there is agreement at least about some of manipulations that can produce those cues. Given this circumstance, the usefulness of the model can be tested. A number of these tests have already been performed and, in my opinion, their results have confirmed several aspects of the model. For example, the results of studies that pitted the memory modulation account of occasion setting against several alternative learning interpretations favored the memory approach. I think many previous accounts of how changes in hunger are linked to changes in appetitive and consummatory behavior have been specified too poorly to yield meaningful assessments of their constructs and of their accounts of the phenomena they are intended to explain.

Sponsored in part by McNeil Specialty Products Co., Merck Research, and the New York Obesity Research Center at St. Luke’s Roosevelt Hospital Center.


SEMINARS AND CONFERENCES || OBESITY RESEARCH CENTER