Organizing Behavior via Motivational States

In the late 19^th century, Darwin suggested that emotions are adaptive in that they serve to prime or trigger a motivated behavioral response [10]. This early thinking is reflected in contemporary cognitive and neuroscientific theories that emphasize functional links between abstract thought and action. For instance, many recent theories [11,12,13,14,15,16] argue that emotional experience is organized by two basic and distinct motivational systems, appetitive and avoidant. The appetitive system is responsible for approach-related behavior, and is activated in response to pleasant and rewarding stimuli. In contrast, the avoidant (or defensive [3]) system is responsible for motivating active avoidance of negative, aversive stimuli (e.g., disgust, punishment, threat [17]). Behavioral repertoire activated by motivational systems can manifest in several ways, such as through the orienting of attention [11,18,19,4], shifting of body postures [9], arm movements [20,21], and even facilitating or inhibiting the onset of locomotion [22,23] toward and away from stimuli. In each case, the direction of movement toward or away from a stimulus results from an hedonic evaluation along the appetitive-avoidant axis; specifically, approach motivations serve to move people closer to an appetitive stimulus, and avoidance motivations do the opposite [3].

Links between motivational states and covert (e.g., postural sway) and overt (e.g., locomotion) actions manifest through activity in a diverse brain network including the amygdala, nucleus accumbens, ventral striatum, orbitofrontal cortex, and anterior cingulate (for a review, see 24). This brain network provides very rapid assessments of motivational relevance. For instance, event-related potentials (ERPs) suggest that the human brain can attach motivational relevance to pictorial stimuli as early as 125-175 ms following stimulus onset [25,26]. There also appears to be emerging evidence that approach and avoidance systems are differentially reliant on the two brain hemispheres, with left hemisphere regions providing support for approach-related states, and right hemisphere for avoidance-related states [27,28]. For instance, asking people to adopt approach-related postures (e.g., leaning forward with arms extended) activates the left hemisphere, and leaning backward produces the opposite neural pattern [29]. Hence, there is evidence for strong links between activating motivational systems and certain movement types, as well as partially independent neural substrates for approach versus avoidance-related thought and behavior.

Eating Behavior and Motivational States

Seminal theories explaining human feeding behavior focused on the regulation of homeostatic mechanisms toward controlling food intake. In general, these theories posited a complex regulatory system that precisely maintains both energy balance and body weight (for reviews, see 30,31. More recently, evolutionary (e.g., [32]), neuroscientific (e.g., [33,34]), and behavioral (e.g., [35,36]) evidence supports the utility of complementing homeostasis-based theory with consideration of appetitive motivational states in accounting for real-world human eating behavior (cf. [37]). For this reason, many current theories attempting to explain human food consumption suggest important interactions between homeostatic and hedonic reward-based mechanisms in motivating attention toward palatable foods and driving eating behavior [38.39].

Recent neuroimaging studies have demonstrated a network of brain regions responsible for processing reward across a variety of stimuli, including food, drugs, and alcohol (e.g., [40]). These include the amygdala and orbitofrontal cortex, two densely interconnected regions that appear to be involved in linking an emotionally salient stimulus to memory representations of reward, and predicting the outcomes of consumption, respectively [41]. These two brain regions work in concert with dopaminergic structures including the striatum and anterior cingulate [42]. Indeed dopamine has been implicated as the primary neurotransmitter involved in signaling reward potential in response to food stimuli [43]. Dysregulation of this neural circuitry, and dopaminergic response, have been implicated in obesity; for instance, Wang and colleagues have demonstrated that obese individuals show a more rapid and pronounced dopamine response than normal weight individuals, during exposure to food images [44]. The endogenous opioid system is also thought to interact with this dopaminergic network and appears responsible for the rewarding properties of specifically palatable foods. Activating this reward system through exposure to preferred food items is responsible for facilitating approach-related behavior [37].

Postural Effects of Motivational State Activation

Bridging the gap between the motivational states and eating literatures, some recent research has demonstrated that overlapping behavioral and neural mechanisms may mediate approach-avoidance behaviors for both food and non-food stimuli. For instance, Gable and Harmon-Jones [45] asked participants to rate their preference for desserts and current hunger, and then exposed them to dessert and neutral pictures while recording brain activity using electroencephalography (EEG). The hungrier the individual was, the greater the left versus right frontal activation while viewing the appetitive desserts versus the neutral images. More recently, Harmon-Jones and colleagues [24] asked participants to adopt forward- or backward-leaning postures while viewing appetitive (dessert) versus neutral stimuli. The approach-related forward lean increased left versus right frontal EEG activity during the viewing of appetitive relative to neutral pictures (see also, 46). Thus, it is becoming clear that body postures can activate lateralized brain regions that are associated with varied motivational states, and these effects can be extended specifically to food-related stimuli.

However, much of the research examining postural variation in response to viewing images selected to promote approach versus avoidance states has found equivocal results. Hillman and colleagues [9] displayed positive, negative and neutral images gathered from a pre-rated image database (IAPS [47]) while monitoring postural sway. Females tended to lean away from unpleasant images (depicting mutilation or attack) relative to pleasant or neutral images, whereas males tended to lean toward pleasant and unpleasant images relative to neutral images. Two additional studies examined postural sway in response to mutilation images and found evidence for reduced anterior–posterior sway, suggesting activation of a defensive freezing posture [48,49]. Similarly, Stins and Beek [50] found no evidence of anterior–posterior sway in response to viewing unpleasant images, but instead found evidence for defensive freezing.

The Present Study

We propose that equivocal postural effects of positive versus negative image viewing are partially due to the highly subjective nature of hedonic evaluation [51,3]. Relying on standardized images may limit the reliability of hedonic evaluations within image conditions; for instance, whereas one participant might find an image of a family dining at a fast food restaurant as positive, another might find it relatively unpleasant. Similarly, though males tend to rate images of erotica as pleasant, females tend to find them relatively neutral or unpleasant [47]. Thus, while exceedingly negative images depicting mutilation might elicit somewhat reliable postural effects (i.e., freezing or anterior-posterior sway), images categorized as pleasant may not reliably activate approach states. In the present study, rather than assuming that particular images are preferred versus non-preferred, we aimed to understand how individuals’ unique preferences modulate postural responses. This type of design affords a more individualized understanding of whether postural sway might reliably index the activation of approach versus avoidance states. We expected that under these conditions, highly preferred foods would elicit anterior-going sway, and highly non-preferred foods would elicit posterior-going sway. To test this hypothesis, we conducted two experiments; first, a pilot Image Ratings Study aimed at stimulus development, and second, a Main Experiment relating preferences to postural sway.

Image Ratings Study Discussion

Our pilot Image Ratings Study examined perceived nutritional and physical attributes for a large set of potential food images, resulting in a set of images that reliably elicited a range of nutritional ratings without confounding visual complexity.

Method

Materials.

We used the 100 food images down-selected from the Image Ratings Study and described in Table 1. Example images are depicted in Figure 1. Participants stood on a Wii balance board that collected center of pressure (COP) data along two (x, y) axes at 16Hz. The Wii balance board is considered a valid apparatus for assessing standing balance, with adequate reliability and sensitivity when compared to professional grade force platforms [56,57]. The device samples center of pressure data at 16Hz (approximately every 62.5 ms). The Wii board has a standing surface of 45 x 26.5 cm, and it was placed on a hard flat surface in the center of the laboratory.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Example food images from the Main Experiment, ranging in nutritional value. Depicted are sliced apples (A), apple pie (B), grilled chicken and vegetables (C), and fried chicken and vegetables (D). Note: Images shown here are for illustrative purposes only due to copyright restrictions.

https://doi.org/10.1371/journal.pone.0072432.g001

A 17″ LCD computer monitor was attached to the top of a large Manfrotto^TM 546B tripod that adjusted vertically to match each participant’s eye height; the monitor was positioned directly in front of the participant, 48″ from the eye. Several demographic variables were collected, including age, height, weight, and gender (BMI was calculated). We also administered a series of questionnaires related to eating behavior. These included the Three-Factor Eating Questionnaire (TFEQ; [58]), the Satiety Labeled Intensity Magnitude (SLIM; [59]), and questions aimed at assessing overall eating habits (poor to excellent), current weight goals (lose, maintain, gain), and history of weight loss programs, and two months’ history of weight fluctuation (lost, maintained, gained).

Preferences and Nutritional Content

A stepwise linear regression examined whether the rated (see Image Ratings Study) healthiness or nutritional content (calories, carbohydrates, fats, sugars, protein) of foods predicted preference ratings. The regression model showed that healthiness was the only factor independently predictive of preference ratings, F(1, 9748) = 922.51, p < .001, R = .29. Overall, participants tended to give higher preference ratings to foods with higher rated healthiness (β_std = .29, p < .001).

Postural Sway Analyses

We found evidence that individual preferences modulated anterior–posterior postural sway, as shown in Figure 2. The ANOVA demonstrated main effects of Preference, F(2,198) = 10.27, p < .001, η² = .06, and Time, F(5,495) = 2.84, p = .015, η² < .01. These effects were qualified by an interaction between Preference and Time, F(10,990) = 10.16, p < .001, η² = .02. Simple effects ANOVAs tested for Preference effects within each Time bin. Time bins 1 (0–500 ms) and 2 (500–1000 ms) did not reach significance (p = .89, p = .09, respectively). Effects were found in Time bins 3 (1000–1500 ms, F(2,198) = 6.12, p < .01, η² = .06), 4 (1500–2000 ms, F(2,198) = 10.46, p < .001, η² = .10), 5 (2000–2500 ms, F(2,198) = 14.14, p < .001, η² = .13), and 6 (2500–3000 ms, F(2,198) = 13.31, p < .001, η² = .12). Participants showed reliable sway toward high preference items and posterior sway in the other two conditions. To ensure that our results were not driven by the unequal number of trials per condition, follow-up analyses using a random sample of trials for each condition equal to the number of trials in the smallest condition (1725), showed maintained significance in the Preference x Time interaction (p < .001), and in Time bins 3 (p = .02), 4 (p < .01), 5 (p < .001), and 5 (p < .01) (cf. [61]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Anterior (forward) versus posterior (backward) postural sway as a function of food item preferences and time relative to image onset. Fixation (referenced to 500ms post-onset baseline) provided for visual comparison.

https://doi.org/10.1371/journal.pone.0072432.g002

To specifically examine these effects, we conducted pair-wise comparisons within time bins 3-6 using a Bonferroni correction term (α_crit = .017). Within time bin 3, the Highly Preferred condition showed more anterior lean relative to the No Strong Preference, t(99) = 2.82, p < .01, d = .42, and Non-Preferred, t(99) = 2.79, p < .01, d = .41, conditions (No Strong Preference vs. Non-Preferred, p = .38). Within time bin 4, the Highly Preferred condition showed more anterior lean relative to the No Strong Preference, t(99) = 3.14, p < .01, d = .42, and Non-Preferred, t(99) = 3.71, p < .001, d = .41, conditions (No Strong Preference vs. Non-Preferred, p = .13). Within time bin 5, the Highly Preferred condition showed more anterior lean relative to the No Strong Preference, t(99) = 3.38, p < .001, d = .52, and Non-Preferred, t(99) = 4.35, p < .001, d = .64, conditions; the Non-Preferred also showed more posterior lean relative to the No Strong Preference condition, t(99) = 2.44, p < .017, d = .35. Finally, within time bin 6, the Highly Preferred condition showed more anterior lean relative to the No Strong Preference, t(99) = 3.53, p < .001, d = .54, and Non-Preferred, t(99) = 4.11, p < .001, d = .62, conditions (No Strong Preference vs. Non-Preferred, p = .04, d = .30).

An ANOVA examining mediolateral (left–right) sway showed a main effect of Time, F(5,495) = 2.91, p = .01, η² < .01, with participants tending to lean slightly more rightward over time. There was no effect of Preference (p = .71), and there was no interaction (p = .11).

A linear regression demonstrated that demographics (age, gender, BMI), eating behavior, weight goals and eating habits did not reliably predict the magnitude of individuals’ anterior/posterior responses to preferred versus non-preferred food items, F(8,98) = 1.87, p = .07, R²_max = .14.

An exploratory ANOVA showed that median splits by rated healthiness did not reliably predict anterior–posterior sway, though participants did show numerical evidence of anterior sway toward relatively healthy, and posterior sway away from relatively unhealthy, foods. This pattern corresponds to regression results.

Limitations

The present college student sample was characterized by relatively healthy eating habits, a restricted BMI range, and low dieting rates. Only 14 of our 100 participants were overweight (BMI range 25-29.9), and four were obese (BMI > 30), according to current recommendations; further, only four participants reported a history of weight loss program enrollment. Similarly, no participants noted anything more than slight hunger during the experimental session. A broader range of eating habits, BMI, dieting history, and state hunger might show value toward predicting postural sway in response to foods. Future research will sample participants from a wider demographic to assess the universality of preference influences on postural sway. A second limitation is related to the limited sensory experience on behalf of participants. Food intake is a multi-sensory experience involving not only vision but also taste, smell, sound, and texture; though we did find compelling results using only vision, more immersive sensory experiences that involve actual food exposure might be expected to elicit more robust postural effects. As Kringelbach [63] notes, the extent of neural response to food exposure might be positively predicted by the nature and extent of sensory input.