Social reward from giving food to others affects food craving and brain potentials: An imagery-based event-related potential study

The interactions between social and eating behaviors can be quite pronounced and are mediated by overlapping neural circuits. The present event-related potential study tested whether the imagery of a specific prosocial behavior (giving chocolates to grateful children) can influence food cue reactivity. A total of 92 females (mean age = 23.5 years) were randomly assigned to one of three guided imagery conditions. The participants listened to an audio recording and were instructed to imagine one of three possible scenes; giving 30 M&Ms to children, eating 30 M&Ms, or sorting 30 marbles. Directly after the imagery task, the participants were presented with images of M&Ms and marbles while their electroencephalogram was recorded. We examined the Late Positive Potential (LPP) across a fronto-central and a parieto-occipital cluster, M&M craving, and subsequent consumption of (real) M&Ms. The mental imagery of offering M&Ms to children was associated with lower M&M craving and higher fronto-central LPP amplitudes (300-600 ms after picture onset) compared to the other imagery conditions. The consumption of M&Ms did not differ between the groups. The LPP is sensitive to the implementation of craving regulation strategies. Furthermore, heightened LPPs are reliably observed in response to motivationally significant stimuli, conflict, and social context. Future studies are needed to specify the specific psychological processes that are associated with the observed LPP effect. In conclusion, this study demonstrated that mental imagery of receiving a social reward from giving food to others can change components of food cue reactivity in healthy females.


Introduction
Food liking, food choices, and eating behavior are strongly influenced by social factors (for reviews see Cruwys et al., 2015;Higgs & Thomas, 2016). For example, in a study by Robinson and Higgs (2012) participants reported reduced liking of orange juice when they believed that their in-group did not like the juice. de la Haye et al. (2013) conducted a longitudinal social network study on friendship and eating behavior of school children. The authors found that the food choices of best friends became or remained similar over the school year. A meta-analysis by Vartanian et al. (2015) that included 38 studies revealed a large social modeling effect on the amount of food intake. The participants ate more when a model (e.g., a live model or remote confederate of the experimenter) ate more and ate less when a model ate less. The effects of the social models were greater in female participants and regarding inhibition of food consumption. The mentioned studies demonstrate that social factors can have a great impact on dietary selection and eating.
Eating and feeding can in turn influence social processes. For example, giving food to others and sharing food in a group strengthens social connections. A study by Woolley and Fishbach (2017) showed that people who ate the same food felt closer to each other and were more trusting. Based on their findings, Miller et al. (1998) concluded that food sharing is an important non-verbal indicator of positive relationships and friendship. Moreover, food offering can be a means to increase positive affect in both the recipient andwhen the offer has the desired effectin the provider (Hamburg et al., 2014).
In the animal kingdom, food sharing is used for mate attraction, sexual access, and mate retention (Alley, 2014). Moreover, feeding is associated with social interaction and bonding with the offspring. It has been argued that offspring feeding is an antecedent of sociality (for a review see Fischer & O'Connell, 2017). Within this context, parents often reduce their food consumption to feed their offspring rather than themselves.
The given examples show that the interaction between social-and eating/feeding behaviors can be quite pronounced. This interaction or overlap can also be seen on a neurobiological level. Neural systems that regulate food intake and social behaviors involve related circuitry (Jennings et al., 2019). In animals and humans, the orbitofrontal cortex holds neurons that are responsive to both social rewards and food rewards. In a study with mice, Jennings et al. (2019) demonstrated that the selective ontogenetic stimulation of social-responsive orbitofrontal neurons reduced the amount of food intake in the presence of a caloric reward (a sweet solution). The authors argued that the brain prioritizes social interaction over food consumption and that the activation of the social circuit inhibited the eating circuit.
The present study investigated a mental imagery strategy to modify attention to visual food cues, food craving, and consumption. The experimental design was similar to an event-related potential (ERP) experiment that tested the effects of imaginary eating to reduce visual food cue reactivity (Zorjan et al., 2020). This approach by Morewedge et al. (2010) had been identified as an effective strategy to reduce the consumption of sugar-coated chocolate candies (M&Ms). Zorjan et al. (2020) administered guided imagery scripts that either described the eating of 30 colorful M&Ms, the sorting of 30 M&Ms by color, or the sorting of 30 marbles by color. Subsequently to the imagery task, the participants were presented with images of M&Ms and marbles while their electroencephalogram and craving ratings were recorded. The results showed that imaginary eating did not reduce the appetitive value of M&M pictures. M&M sorting even enhanced craving and late positivity toward M&M pictures (300-600 ms after picture onset) across a parieto-occipital cluster.
Therefore, we tested a new strategy to reduce food cue reactivity. We asked the participants to imagine how they engage in food-related prosocial behavior. More specifically, the participants were asked to imagine giving sweets (M&Ms) to children, who take the little chocolates happily and gratefully. Thus, the participants received a social reward for not eating the sweets themselves and feeding the children. Subsequently, the participants were presented with images of M&Ms (and marbles as control stimuli).
As a neurobiological measure of attention directed to the food cues, we focused on ERP late positivity as in the previous study by Zorjan et al. (2020). Components of late positivity such as the P300 and the Late Positive Potential (LPP) reflect the motivational relevance of stimuli (for a review see Olofsson et al. (2008). The P300, a positive deflection in the EEG that starts around 300 ms post-stimulus onset, has been linked with automatic attention allocation to stimuli with intrinsic motivational significance, whereas the subsequent LPP (which can last up to 6000 ms) reflects more deliberate attentional processes and is sensitive to emotion regulation strategies Hajcak et al., 2010;Schupp et al., 2000Schupp et al., , 2004. EEG research has demonstrated that P300/LPP amplitudes are greater for images of food vs. non-food (Blechert et al., 2014;Nijs et al., 2010;Sarlo et al., 2013;Stockburger et al., 2009;Zorjan et al., 2020). Moreover, P300 and/or LPP amplitudes to food images have been positively correlated with self-reports of craving and food intake (Biehl et al., 2020;Nijs et al., 2010).
Based on these findings, we hypothesized that the social reward associated with offering M&Ms to children should reduce late positivity (P300/LPP amplitudes) to M&M images and the craving/consumption of M&Ms.

Participants
We used G*power (Faul et al., 2007) to estimate the sample size for the present investigation. To detect a small effect size and achieve a power of .80 or above at the alpha level of 0.05, we would need a sample of at least 87 participants. We recruited a total of 97 participants to account for possible exclusions due to artifacts. All participants were university students who were recruited via campus announcements and social network advertisements. Five participants were removed from the analysis due to excessive artifacts in the EEG. The remaining 92 females had a mean age of 23.50 years (SD = 3.30 years) and reported a mean BMI of M = 22.0 (SD = 4.0). Exclusion criteria were reported current/previous somatic (neurological) illness (e.g., food allergies, head injuries) and mental disorders (e.g., eating disorders). We only included female participants because of gender differences concerning reported cravings. Several studies have shown that males report more cravings for savory foods, whereas females report more cravings for sweet foods (e. g., Weingarten & Elston, 1991).
The participants were randomly assigned to one of three mental imagery groups: M&M Eating (n = 29), M&M Giving (n = 33), and Marble Sorting (n = 30). Before the experiment, all participants reported their hunger level (M = 3.80, SD = 2.40) and general liking of M&Ms (M = 6.70, SD = 1.40) on scales from 1 (not hungry at all; no liking) to 9 (very hungry; strong liking). Mean hunger level, age, BMI, and M&M liking did not differ between the groups (all p > .14).

Imagery scripts
The participants listened to a 10-min audible recording (via loudspeakers; female voice) that either described the giving of 30 M&Ms to children, the eating of 30 M&Ms, or the sorting of 30 marbles. The imagery instructions for eating M&Ms/sorting marbles were the same as used by Zorjan et al. (2020). The participants were asked to close their eyes and to imagine that they were sitting in a comfortable chair beside a table with a white bowl that was filled with colorful M&Ms or with colorful marbles. In the M&M Giving condition, participants were asked to repeatedly reach into the bowl, take an M&M of a certain color, and hand the M&M to a child that took the little chocolate happily and gratefully (e.g., the child smiled, said 'Thank you so much.'). In the M&M Eating condition, participants were asked to repeatedly reach into the bowl, take M&Ms of certain colors, and eat them. In the Marble Sorting condition, participants were asked to reach into the bowl, take out marbles of certain colors, and sort them according to their color. The three conditions were identical concerning the beginning (introduction) and conclusion of the imagery exercise and had a comparable duration (word count).

Images
The participants viewed a total of 60 different pictures from two categories: M&Ms (30 pictures) and marbles (30 pictures). The pictures were shown in randomized sequence for 3000 ms each on a black background and were preceded by a white fixation cross on a black background (500-1000 ms). During the picture presentation, participants were asked five times to rate their current M&M craving using a 9point Likert scale ranging from 1 (low) to 9 (high). The craving questions were randomly distributed throughout the task. The images and the presentation design were the same as in the study by Zorjan et al. (2020). Images and imagery scripts are provided at OSF (open science framework; https://osf.io/grf6w/).

Procedure
Participants first provided written informed consent and completed an online screening to check for inclusion/exclusion criteria. Those participants who were eligible to participate were invited to the EEG laboratory. They were instructed to refrain from eating for at least 3 h before attending the laboratory session. After arriving at the lab, EEG electrodes were placed. The participants were instructed and completed a test trial. The test trial consisted of the presentation of one image depicting M&Ms (not later used in the experiment) followed by the craving rating (slider style rating). The test trial was conducted to familiarize the participants with the rating procedure. Subsequently, they listened to the guided imagery instruction (∿ 10 min), which was followed by the picture presentation (∿ 10 min) during which the EEG was recorded. At the end of the experiment, participants completed a filler questionnaire (where they provided information on their semester, major field of study, interest in research credits for participation). Additionally, they were given a bowl containing 500 g of M&Ms. They were told that they could eat as many M&Ms as they wanted during the completion of the questionnaire (fixed 10 min-interval). The amount of M&Ms eaten (in grams) was measured. The study was carried out following the Declaration of Helsinki and was approved by the ethical committee of the university.

Electrophysiological recording and data analyses
Data were recorded with an actiCHamp system (actiCHamp, Brain Products GmbH, Gilching, Germany) using 63 active actiCAP snap electrodes (according to the 10-10 system) and the BrainVision Recorder (version 1.21). The reference electrode was placed on position FCz, the ground electrode on position FPz. An electrolyte gel was applied to each electrode to keep electrode impedances below 10 kΩ. The EEG was recorded with a sampling rate of 2500 Hz and a passband of 0.016-1000 Hz. For raw data analysis, the BrainVision Analyzer (version 2.0.4) was used. The sampling rate was changed to 250 Hz. The data were rereferenced to linked mastoid electrodes (i.e., TP9, TP10). Artifacts due to eye movements were corrected via the implemented ICA ocular correction softwareonly components corresponding to horizontal and vertical eye movements were selected based on the correspondence of their shape, timing, and topography. Further artifact episodes were excluded after visual inspection. Five participants were excluded from the analysis due to a large number of artifacts (<70% of artifact-free segments). For the remaining participants, the percentage of artifactfree trials did not differ between groups ( Data were segmented in 3200 ms intervals (200 ms pre-stimulus onset, 3000 ms post-stimulus onset) and corrected to the 200 ms prestimulus baseline. An offline high-pass (0.1 Hz) and low-pass filter (cut-off frequency 30 Hz, roll-off 24 dB/octave) were applied. Data were averaged for all groups and conditions separately.

Statistical analysis
We computed mixed 2x3 ANOVAs to test the effects of GROUP (between-subjects: imaginary M&M Eating, M&M Giving, Marble Sorting) and PICTURE (within-subjects: M&Ms, Marbles) on LPP amplitudes. Group differences concerning M&M craving and consumption were tested with one-way ANOVAs. Significant main and interaction effects in the ANOVAs were followed up with t-tests. We report adjusted p values using the Bonferroni-Holm method. As a measure of effect size, we report partial eta squared for the ANOVA analyses and Cohen's d for the follow-up t-tests.

Exploratory follow-up analyses for P200
Based on the suggestions of a reviewer, we conducted additional analyses for an earlier ERP component (P200: 180-250 ms). Fronto

Exploratory correlation analyses
The reported M&M craving correlated positively with M&M consumption (r = 0.37, p < .001). The frontal/parietal LPP amplitudes for M&M images showed no significant correlations with craving and M&M  consumption (all p > .14).

Discussion
This ERP study investigated an imagery-based approach to change food cue reactivity. The participants imagined that they offered sugarcoated chocolate candies to children and obtained social rewards for this action (e.g., smiles, verbal expressions of gratitude). After the imagery exercise, the reported craving for M&Ms was lower in the M&M Giving group than in the M&M Eating group. The Giving group reported a comparable craving level as the group with no imaginary food exposure (marble sorting).
The lower craving in the Giving group was accompanied by an increased magnitude of the early LPP to M&M pictures across a frontocentral cluster. The LPP has been associated with a range of processes, including the coding of motivational salience (e.g., Hajcak et al., 2010), cognitive control/emotion regulation (e.g., Hajcak & Nieuwenhuis, 2006), integration of social context information (e.g., Schwab et al., 2017), and experience of conflict (Field et al., 2016).
A common interpretation of the LPP is that of an indicator of motivational salience (Hajcak et al., 2010). Stimuli that prompt appetitive or defensive behaviors (e.g., food, threat) lead to enhanced P300/LPP amplitudes. For example, event-related late positivity is greater for images of food vs. non-food (Blechert et al., 2014;Sarlo et al., 2013). The maximal amplitude of this 'motivated attention' effect is typically present across parietal regions. It has been suggested that the temporary increase in motivated attention serves to facilitate the visual processing of the stimulus. The 'enhanced perception hypothesis' of the LPP is consistent with the neural sources of this component in the visual association cortex (Liu et al., 2012). In the present experiment, the LPP showed the expected food-bias effect. All participants displayed higher parietal LPPs to M&M pictures compared to marble pictures. However, we did not detect group differences concerning the parietal LPP. This implies that the imagery conditions did not differentially influence the attention-capturing quality of the presented food cues.
The LPP not only reflects the motivational salience of a stimulus but is sensitive to manipulations that alter the meaning attributed to the stimulus (Hajcak & Nieuwenhuis, 2006). More specifically, in the context of food-cue processing, the LPP amplitude can be altered via craving regulation (Meule et al., 2013;Svaldi et al., 2015). In a study by Meule et al. (2013), the participants (healthy, normal-weight females) were instructed to think about the long-term effects of eating high-calorie foods. This intervention reduced craving and increased LPP amplitudes across a centro-parietal cluster. In contrast, Svaldi et al. (2015) found that compared to passive viewing, engaging in craving regulation (reappraisal or suppression) while watching food pictures reduced the centro-parietal LPP. This result was obtained for a group of restraint eaters.
The interventions described were very likely associated with different types of emotion/mood induction. For example, thinking about the negative consequences of food consumption (Meule et al., 2013) typically leads to a negative affective state with increased arousal. Increased arousal however is also possible in the context of positive emotion. Some participants of the Giving group stated that the mental imagery of the little children taking the chocolates was 'heart-warming' and 'gratifying'. Hypothetically, these feelings could have been associated with higher levels of arousal than the other two conditions. If the participants re-experienced the arousal while viewing the M&M pictures because they were reminded of the experience, this could explain increased LPP amplitudes to M&M pictures in the Giving condition. Future studies should assess possible changes in valence/arousal induced by the imagery scripts. Mood states have been shown to modulate LPPs to food pictures before, particularly in females with different types of dysfunctional eating behaviors (restrained/emotional/bulimic eating; Blechert et al., 2014;Lutz et al., 2021;Schnepper et al., 2020).
The LPP can also be influenced by social context, which has been associated with amplitude modulations across frontal clusters (Foti & Hajcak, 2008;MacNamara et al., 2009;Woodcock et al., 2013). For example, in a study by Woodcock et al., (2013), the presence of different social partners during the viewing of affective pictures modulated frontal LPP amplitudes. Studies by Foti and Hajcak (2008) and Mac-Namara et al. (2009) demonstrated that contextual information provided about affective images altered the associated emotional experience and the frontal LPP. The participants first listened to a brief description of the upcoming picture. Before the presentation of unpleasant images, this description was either more neutral or more negative. Following the neutral description, the magnitude of the LPP, ratings for arousal, and unpleasantness were all lowered. Thus, the frontal LPP reflects socio-cognitive control mechanisms involved in the modulation of emotional experience and expressions (Woodcock et al., 2013) The mentioned findings demonstrate that the LPP is associated with a variety of psychological processes. However, it is difficult to determine which ones were involved in the effects reported in the present study. Of note, LPP modulations were restricted to the early LPP time window (300-600 ms after picture onset) but were not present later on (600-3000 ms). Based on principle component analysis, Foti et al. (2009) have suggested that the LPP can be separated into an early (300-600 ms) and a later component (>600 ms). Functionally, the early LPP is implicated in affective encoding, motivated attention, and arousal, while the late LPP is thought to reflect mnemonic aspects of emotional processing and prolonged effortful processing. As mentioned before, the Giving condition of the present investigation very likely elicited more positive affect and/or arousal than the Eating and Sorting conditions. Moreover, some authors (Glazer et al., 2018) have pointed out that specific components of ERP late positivity can reflect two different aspects of reward processing: reward anticipation vs. reward-outcome evaluationtwo largely independent components. For example, the feedback-related P300 (FB-P300) is a positive deflection peaking from 300 to 600 ms following a feedback stimulus. The FB-P300 involves an attention-driven categorization of salient outcome-related information, such as context updating. For the present experiment, it can be assumed that the Giving condition modified the outcome evaluation for the M&M pictures. Here, the participants not only appraised the food reward of the M&Ms ('The M&Ms would taste good.') but also their response ('I gave the M&Ms to the children and made them happy.'). This additional social reward assignment could have increased the frontal early LPP to M&Ms.
The imagined offering of M&Ms to the children did not influence M&M consumption relative to the other conditions. The average amount eaten was small (13 g) and some participants even did not eat at all. Comparable findings have been reported in an EEG study with a similar design (Zorjan et al., 2020). Here, participants also had the free choice to eat. Other studies on guided imagery effects on M&M consumption prompted the eating by utilizing a bogus taste test (Morewedge et al., 2010), which reduces non-responding rates. Finally, the representativeness of eating in the laboratory can be questioned.
The present study failed to replicate that imaginary eating is a useful approach to reliably reduce craving in a sample of healthy females (Zorjan et al., 2020). Repeated exposure to food cues tends to even increase craving and the risk of overeating (see meta-analysis by Boswell & Kober, 2016). However, imagery of social reward while giving food to others reduced the desire to eat. Therefore, this strategy should now be tested in clinical groups, such as patients with binge-eating symptoms or emotional eating. A study by Cardi et al. (2019) successfully reduced food intake in females with bulimia nervosa and binge eating disorder by positive mood induction. Following exposure to a video designed to induce food craving, a positive mood vodcast (positive classical music along with positive spoken statements) was associated with less food consumption in a taste test meal than a neutral vodcast. Therefore, it seems promising to investigate whether similar effects can be obtained with positive mood induction through social reward.
We also need to mention the following limitations of the present investigation. We studied a female sample of university students. Therefore, the results cannot be generalized to other groups. We did not assess M&M craving before the imagery condition but only afterward. The reported group differences in craving may be due to different processes: the imaginary giving of M&Ms to the children could have reduced the craving, or the imaginary M&M eating could have increased the craving. Moreover, the imagery instruction could be personalized (e. g., giving food to friends and family) to make this approach more effective (e.g., in terms of craving/LPP differentiation between the groups). Finally, the introduction of an additional control condition 'the giving of marbles to grateful children could help to differentiate effects of social reward obtained in the context of food vs. non-food cues on food cue reactivity.
In conclusion, this study demonstrated that imagery of a social reward can change subjective as well as electrocortical components of food cue reactivity in a group of healthy female students.

Data availability
The raw data that support the findings of this study are available from the corresponding author upon reasonable request.

Funding
The research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author contributions
A.S. designed the study. S.Z. and A.G. collected and analyzed the data. All authors were involved in writing the manuscript.

Ethics statement
All participants provided written informed consent.