Reassessing wanting and liking in the study of mesolimbic influence on food intake

Abstract

Humans and animals such as rats and mice tend to overconsume calorie-dense foods, a phenomenon that likely contributes to obesity. One often-advanced explanation for why we preferentially consume sweet and fatty foods is that they are more "rewarding" than low-calorie foods. "Reward" has been subdivided into three interdependent psychological processes: hedonia (liking a food), reinforcement (formation of associations among stimuli, actions, and/or the food), and motivation (wanting the food). Research into these processes has focused on the mesolimbic system, which comprises both dopamine neurons in the ventral tegmental area and neurons in their major projection target, the nucleus accumbens. The mesolimbic system and closely connected structures are commonly referred to as the brain's "reward circuit." Implicit in this title is the assumption that "rewarding" experiences are generally the result of activity in this circuit. In this review, I argue that food intake and the preference for calorie-dense foods can be explained without reference to subjective emotions. Furthermore, the contribution of mesolimbic dopamine to food intake and preference may not be a general one of promoting or coordinating behaviors that result in the most reward or caloric intake but may instead be limited to the facilitation of a specific form of neural computation that results in conditioned approach behavior. Studies on the neural mechanisms of caloric intake regulation must address how sensory information about calorie intake affects not just the mesolimbic system but also many other forms of computation that govern other types of food-seeking and food-oriented behaviors.

Keywords: liking; reward; wanting.

Fig. 1.

Three forms of approach behavior. The diagrams show three forms of approach behavior that require different neural computations (187, 198, 272, 273). Start and target locations are represented by the beginning and end of arrows, respectively. Cognitive map-based navigation: when the target object cannot be directly sensed, but its location with respect to distal stimuli is known, approach is based on a cognitive map constructed from the spatial relationships among these stimuli. Taxic approach: when the target object is visible, approach is guided by the object. At each starting point, the brain must determine a novel set of actions to reach the target. Praxic approach: the same actions are used each time a given object is approached. This mode of approach is possible only when the start and target locations are always the same relative to each other across instances of approach.

Fig. 2.

Behavior and nucleus accumbens neuronal firing during rats’ performance of a discriminative stimulus task. A: video tracking data taken from an example session show the rat’s position at discriminative stimulus (DS) onset within the operant chamber (dots). The DS indicates the availability of a droplet of sucrose (in the receptacle) contingent on a lever press. The intertrial interval (ITI) is variable and averages 30 s. The rat moves about the chamber during the ITI. Therefore, its starting location with respect to the approach target (the lever) is variable, and it must engage in taxic approach to earn sucrose during DS presentation. Infusions of dopamine receptor antagonists into the nucleus accumbens (NAc) impair the ability of animals to initiate taxic, but not praxic, approach (not shown). (Republished from Ref. .) B: an example neuron demonstrating the properties of the ~45% of NAc neurons that are excited by presentation of the DS. The rasters (top) show the time at which this neuron fires an action potential, aligned to the DS onset in each trial. The rasters are sorted by the latency to reach peak velocity (blue dots), which occurs shortly after the initiation of approach movement. The histograms (bottom) show the average firing rate in the quartiles of trials with shortest and longest latency. (From data reported in Ref. .) C: histograms showing average firing, aligned to DS onset, across NAc DS-excited neurons. Antagonists of dopamine D1 (left) and D2 (right) receptors were infused into the NAc after a 45-min baseline during which the animal performed the task. (Republished from Ref. .)

Fig. 3.

Influence of the animal’s proximity to the movement target on behavior and accumbens neuronal firing during rats’ performance of a choice task. A: in the choice task, the cues indicating availability of a choice (a light and extension of two levers) were presented noncontingently in an unpredictable way. One lever was always optimal (it delivered more sucrose reward than the other), and the optimal lever changed across blocks. Choice blocks were preceded by forced trials (one lever at a time was presented) to teach the animal how much sucrose was associated with each lever. B: animals’ locations at cue onset were variable, sometimes close to the left lever (1), close to the right lever (2), or far from both levers. C: choice was impacted by proximity. Animals tended to make more suboptimal choices when they were close to the suboptimal lever than when they were close to the optimal lever or far from the levers. D: average cue-evoked firing of NAc neurons was greater when the animal was close to a lever than when far from the levers. Note that the average firing response in a given proximity condition was nearly identical when large and small reward levers were chosen. This result indicates that firing strongly encodes proximity of a movement target but not predicted outcome value. L lever, left lever; R lever, right lever; lg, large reward; sm, small reward. (Republished from Ref. .)