Hemispheric asymmetries in resting‐state EEG and fMRI are related to approach and avoidance behaviour, but not to eating behaviour or BMI

Abstract Much of our behaviour is driven by two motivational dimensions—approach and avoidance. These have been related to frontal hemispheric asymmetries in clinical and resting‐state EEG studies: Approach was linked to higher activity of the left relative to the right hemisphere, while avoidance was related to the opposite pattern. Increased approach behaviour, specifically towards unhealthy foods, is also observed in obesity and has been linked to asymmetry in the framework of the right‐brain hypothesis of obesity. Here, we aimed to replicate previous EEG findings of hemispheric asymmetries for self‐reported approach/avoidance behaviour and to relate them to eating behaviour. Further, we assessed whether resting fMRI hemispheric asymmetries can be detected and whether they are related to approach/avoidance, eating behaviour and BMI. We analysed three samples: Sample 1 (n = 117) containing EEG and fMRI data from lean participants, and Samples 2 (n = 89) and 3 (n = 152) containing fMRI data from lean, overweight and obese participants. In Sample 1, approach behaviour in women was related to EEG, but not to fMRI hemispheric asymmetries. In Sample 2, approach/avoidance behaviours were related to fMRI hemispheric asymmetries. Finally, hemispheric asymmetries were not related to either BMI or eating behaviour in any of the samples. Our study partly replicates previous EEG findings regarding hemispheric asymmetries and indicates that this relationship could also be captured using fMRI. Our findings suggest that eating behaviour and obesity are likely to be mediated by mechanisms not directly relating to frontal asymmetries in neuronal activation quantified with EEG and fMRI.

eating behaviour and BMI. We analysed three samples: Sample 1 (n = 117) containing EEG and fMRI data from lean participants, and Samples 2 (n = 89) and 3 (n = 152) containing fMRI data from lean, overweight and obese participants. In Sample 1, approach behaviour in women was related to EEG, but not to fMRI hemispheric asymmetries. In Sample 2, approach/avoidance behaviours were related to fMRI hemispheric asymmetries. Finally, hemispheric asymmetries were not related to either BMI or eating behaviour in any of the samples. Our study partly replicates previous EEG findings regarding hemispheric asymmetries and indicates that this relationship could also be captured using fMRI. Our findings suggest that eating behaviour and obesity are likely to be mediated by mechanisms not directly relating to frontal asymmetries in neuronal activation quantified with EEG and fMRI.

K E Y W O R D S
approach/avoidance behaviour, BMI, EEG, fMRI, hemispheric asymmetries, obesity, restingstate 1 | INTRODUCTION A sizeable proportion of our everyday actions is driven by approach (e.g. reaching for a tasty biscuit) and avoidance (e.g. running away from a big spider) tendencies. Such tendencies can be considered fundamental motivational dimensions that steer (not only) human behaviour (Davidson & Hugdahl, 1995). These two dimensions are at the core of the framework of behavioural inhibition and activation systems (BIS and BAS, respectively; Gray, 1981;Gray & McNaughton, 1992) and can, for example, be assessed by means of the self-report BIS/BAS questionnaire (Carver & White, 1994). Literature on individual differences in terms of inhibition and activation systems is broad and mostly focuses on disorders such as depression, anxiety, substance addictions, or obesity (Dietrich, Federbusch, Grellmann, Villringer, & Horstmann, 2014;Johnson, Turner, & Iwata, 2003;Morgan et al., 2009). There is experimental evidence that both substance addictions and obesity are related to increased approach behaviour towards problematic stimuli: While substance abuse relates to approach towards cigarettes, marijuana, or alcohol substances, obesity relates to approach tendencies towards unhealthy food cues (Cousijn et al., 2012;Mehl, Morys, Villringer, & Horstmann, 2019;Mehl, Mueller-Wieland, Mathar, & Horstmann, 2018;Wiers et al., 2013;Wiers et al., 2014). Furthermore, obesity and higher body mass index (BMI) were shown to relate to BIS/BAS scores in a gender-dependent fashion, with positive correlations in women, and negative correlations in men (Dietrich et al., 2014).
Regarding the neural correlates of approach/avoidance behaviours, literature suggests differential engagement of left and right frontal brain areas, such as the Brodmann area 9 or 10, and rewardrelated regions of the brain, such as the nucleus accumbens or the ventral tegmental area (Aberg, Doell, & Schwartz, 2015;Tomer et al., 2013). The left hemisphere is more strongly engaged in approach, while the right one in avoidance behaviours (Aberg et al., 2015;Davidson, 1993Davidson, , 1994Sutton & Davidson, 1997;Tomer, Goldstein, Wang, Wong, & Volkow, 2008). A seminal study showed that higher alpha power, which is believed to represent inhibitory control (Bazanova & Vernon, 2014;Klimesch, Sauseng, & Hanslmayr, 2007), in right frontal brain areas (relative to the left) measured in restingstate EEG (rsEEG), was associated with increased approach behaviour (Sutton & Davidson, 1997). This was explained by downregulated right hemispheric activity since alpha power has previously been linked to cortical inhibition by top-down control and suppression of task-irrelevant brain regions (Bazanova, 2012;Klimesch et al., 2007).
A number of studies showed similar functional asymmetries in reward regions such as the ventral tegmental area and nucleus accumbens using positron emission tomography (Tomer et al., 2013) and taskbased fMRI (Aberg et al., 2015) during reward and punishment learning. These findings suggest that hemispheric asymmetries and their relationship to approach/avoidance behaviours can be quantified using a range of neuroimaging tools. However, the relationship between approach/avoidance behaviours and hemispheric asymmetries in resting-state fMRI (rsfMRI) has not yet been investigated.
Since obesity is related to altered approach/avoidance behaviours, it might also be related to hemispheric asymmetries. This hypothesis is grounded in the right-brain theory of obesity, which posits that hypoactivation of the right prefrontal cortex is an underlying factor of obesity (Alonso-Alonso & Pascual-Leone, 2007). It is based on findings of increased eating behaviour after damages to right-hemispheric anterior brain areas (Regard & Landis, 1997;Short, Broderick, Patton, Arvanitakis, & Graff-Radford, 2005). It is also supported by EEG experiments showing a higher right-hemispheric bias for restrained eaters, a predominantly inhibitory feature (Silva, Pizzagalli, Larson, Jackson, & Davidson, 2002) and a positive relationship of left-hemispheric bias with disinhibition and hunger (Ochner, Green, van Steenburgh, Kounios, & Lowe, 2009) as measured with the three-factor eating questionnaire (TFEQ; Stunkard & Messick, 1985).
In this study, we addressed three aims using three independent samples. First, we aimed to conceptually replicate the previous find-  (Beck, Steer, Ball, & Ranieri, 1996). Additional data available for this sample were self-reported eating behaviour (TFEQ) data and anthropometric data (BMI ;   Table S2).
With regard to the self-reported eating behaviour, we used the three-factor eating questionnaire (TFEQ; Stunkard & Messick, 1985). 2.4.2 | fMRI data-Samples 1 and 2 fMRI data pre-processing for Samples 1 and 2 were identical and was done within the Nipype framework (Gorgolewski et al., 2011). In short, the pre-processing steps included discarding the first five func-

| fMRI data-Sample 3
fMRI data pre-processing for Sample 3 data was also done within the Nipype framework (Gorgolewski et al., 2011). In short, the pre- In this step, we attempted to directly replicate previous findings from Sutton and Davidson (1997) showing a positive correlation of lefthemispheric bias with BAS-BIS differential scores. Since this measure is not recommended by authors of the BIS/BAS questionnaire (Carver & White, 1994), we used it in our study only to replicate previous findings of Sutton and Davidson (1997). In this first analysis and in this analysis only, we calculated an absolute EEG asymmetry index in frontal areas by subtracting absolute alpha power (8-12 Hz) in the F3 electrode (left) from absolute alpha power in the F4 electrode (right; asymmetry index: R-L) for mean values of EO and EC conditions together.
We then wanted to extend previous findings concerning EEG hemispheric bias and approach/avoidance behaviour to eating behaviour (as measured by the TFEQ). As rsfMRI was collected with eyes open to prevent subjects from falling asleep, our main analysis focused on EEG data from the eyes open condition in order to compare it with fMRI findings. We additionally conducted EEG analyses with relative alpha power of eyes closed condition to investigate whether potential effects observed in the eyes open condition are specific to this condition or can be extended to the eyes closed condition as well.
While the broader alpha frequency band (8-12 Hz) has been previously linked to cortical inhibition by top-down control (Bazanova, 2012;Klimesch et al., 2007), low alpha power (8-10 Hz) was previously shown to reflect general attentional demands, basic alertness, vigilance and arousal (Klimesch et al., 2007;Petsche, Kaplan, von Stein, & Filz, 1997). Including both of the measures allowed us to replicate previous results obtained using broadband alpha, and confine possible mechanistic interpretations to, for example, general attentional demands (by using low alpha). For this analysis, as opposed to the direct replication described in the previous paragraph, we used relative alpha power to control for inter-individual differences in contaminating factors like skull thickness and meninges that might affect tissue conductivity and influence electrical signal captured at the sensor level (Babiloni et al., 2011). Relative power in broadband alpha and low alpha frequency ranges were calculated by firstly taking the mean of the squared amplitude obtained after filtering the signal in the 8-12 Hz and the 8-10 Hz frequency ranges, respectively, and then dividing it by the power within the frequency range of 4-40 Hz. In line with Sutton and Davidson (1997), relative alpha power measures were calculated in the pair of frontal electrodes F4 and F3. We also used pairs of F5/F6 and F7/F8 electrodes to extend our investigations according to current trends (Harmon-Jones & Gable, 2018). Moreover, we included a parietal pair, P4 and P3, as a control to investigate whether the observed relationship with frontal asymmetries was topographically specific.
Previous research on hemispheric asymmetries used an absolute asymmetry index (Sutton & Davidson, 1997), while in our study we calculated a relative asymmetry index using the following equation: (R − L)/(R + L). By accounting for inter-individual differences in alpha power magnitude, these relative indices capture asymmetries better than the absolute R − L difference and increase interpretability (Hiroshige & Dorokhov, 1997;Pivik et al., 1993). After calculation of asymmetry indices, we excluded outliers from all variables of interest using the a priori defined criterion (see section 2.6). EEG analysis for different electrodes pairs included different numbers of participants due to artefactual channels or outlier exclusions that were performed separately for each variable. We used such strategy to maximise the statistical power of our analyses.
2.5.2 | Aim 2 + 3: Hemispheric asymmetries in fMRI After pre-processing (sections 2.4.2 and 2.4.3), analysis of fMRI data in all three samples was identical. To be able to conceptually compare EEG results with fMRI results, the fractional amplitude of lowfrequency fluctuations (fALFF) was used as a measure of resting-state brain activity (Zou et al., 2008). fALFF is usually defined as the ratio of power in the frequency range of 0.01-0. These ROIs were defined using pickatlas (Maldjian, Laurienti, Kraft, & Burdette, 2003). Since fMRI allows to investigate subcortical brain areas, for which hemispheric asymmetries have been shown ( and Neto, Oliveira, Correia, and Ferreira (2008); L: x = −9, y = 9, z = −8; R: x = 9, y = 8, z = −8

| Statistical analysis
For each of the variables of interest, outliers were excluded based on an a priori criterion: 2.2*interquartile range below or above the first or third quartile, respectively (Hoaglin & Iglewicz, 1987;Hoaglin, Iglewicz, & Tukey, 1986;Tukey, 1977  Relationships between fMRI hemispheric asymmetries and approach/ avoidance and eating behaviours in Sample 1 To investigate relationships of fMRI hemispheric bias with approach/ avoidance and eating behaviours, we first used rotated principal component analysis (PCA) on the ROI imaging data (asymmetry indices calculated for mean fALFF values per ROI). This was done to reduce the number of comparisons in further analyses (Jolliffe & Cadima, 2016). We used the varimax rotation, which drives component loadings (correlations of components and original variables) either towards zero or towards a maximum possible value, decreasing a number of components with medium loadings, which are difficult to interpret (Jolliffe, 2002;M. B. Richman, 1986;M. L. B. Richman, 1987). As a criterion for retaining components, we chose the minimum cumulative variance explained to be over 70% (Jolliffe, 2002). This resulted in six components for each of the samples.
Furthermore, to investigate relationships of fMRI hemispheric bias and approach/avoidance behaviour, we performed a similar analysis to the one using EEG data. Six rotated principal components were defined as outcome measures, and predictors included BAS fun, BAS drive, BAS reward responsivity, as well as BIS anxiety and FFFS fear scores and their interaction with gender. Additionally, we included BMI and age as variables of no interest (Bonferroni corrected α = .0084, n = 110).
A similar analysis was performed to investigate relationships between fMRI hemispheric bias and eating behaviour. It included similar predictors as the EEG investigation of eating behaviour-cognitive control and disinhibition and their interaction with gender. Outcome variables were six rotated principal components. We added BMI and age as variables of no interest (Bonferroni corrected α = .0084, n = 106).

| Aim 1: EEG replication analysis-Sample 1
In this analysis, we aimed to directly replicate findings of Sutton and Davidson (1997) of increased hemispheric bias (R − L; F4 − F3 electrodes, absolute alpha power, mean values for EO and EC conditions) being related to increased BAS − BIS differential scores. We did not find a significant relationship between those variables (r (113) = .121, p = .202). Partial correlation after controlling for BMI, age and gender also did not reveal a significant relationship (r (113) = .094, p = .325).
F I G U R E 1 Relationship between low/full alpha EEG asymmetry index (AI) and BAS drive scores. Index used: (R − L)/(R + L). Triangles/ dots represent data points, dashed/bold lines represent the best fit and grey shaded areas are 95% confidence intervals. (a) Significant correlation of hemispheric asymmetries and behavioural measures in the low alpha spectrum (beta: −.85, p = .0020); (b) not significant correlation of hemispheric asymmetries and behavioural measures in the broad alpha spectrum showing that the asymmetries are specific to the low alpha spectrum (beta: −.14, p = .5476). AI, asymmetry index; L, left; R, right Next, we attempted to expand previous findings linking EEG and approach/avoidance behaviours to (a) additional frequency ranges to improve specificity and interpretability of findings, (b) additional questionnaire measures to improve specificity of the findings. We therefore investigated relationships between EEG parietal and frontal asymmetry indices as measured by the relative broad alpha power, as used by Sutton & Davis, and relative low alpha power. In addition to the standard broad alpha power spectrum used in previous studies, low alpha power spectrum due to its specific physiological meaning (general attentional demands, basic alertness, vigilance and arousal; Klimesch et al., 2007;Petsche et al., 1997) allowed us to more precisely interpret relationships between hemispheric asymmetries and behaviour. Here, we used the improved, relative asymmetry index: (R − L)/(R + L). For questionnaire data, we included BAS fun-seeking, drive, reward responsivity, BIS anxiety and FFFS fear scales. First, we investigated the eyes open condition. Results of this analysis (Table 1) indicate a significant positive relationship of BAS drive and left frontal hemispheric bias in low alpha frequency for women only (BAS drive: p = .0009, BAS drive * gender: p = .0020). This is shown by an interaction of BAS drive with gender, and a significant main effect of BAS drive. In this analysis, women were coded as 0 and were the reference category, hence the main effect of BAS drive shows that this relationship is true for women, because in this case all other interaction terms including gender are also equal to zero. A similar relationship was not significant for broad alpha power. For scatter plots of these relationships see Figure 1. Even though we performed outlier exclusion prior to the analysis, we visually identified data points that could potentially be outliers and hence influence the results (points above 3 and below −3 on the Y-axis, Figure 1a). Removal of these data points, however, did not alter the results. In the analysis of the eyes closed condition we found no significant effects (Table S3). We performed a linear mixed effect model analysis with subject as a random factor and con- 3.2 | Aim 2: fMRI correspondence analysis-Sample 1 First, we investigated direct relationships between EEG asymmetries (using the relative asymmetry index (R − L)/(R + L)) and whole-brain fALFF asymmetry measures in the same sample. This analysis did not produce significant results, suggesting no correspondence between rsEEG and rsfMRI hemispheric bias measures.
Next, we investigated relationships between fMRI relative asymmetry indices (L − R)/(L + R) and approach/avoidance behaviours in Sample 1. The analysis included six retained components describing asymmetry data and questionnaire variables-BAS fun, BAS drive, BAS reward responsivity, BIS anxiety and FFFS fear and their interactions with gender.
Additionally, we included BMI and age as covariates of no interest. We found no significant associations for this analysis (Tables 2 and 3, n = 110).
Furthermore, we investigated whether hemispheric asymmetries measured with fMRI are related to self-reported eating behaviour (TFEQ). This analysis included cognitive control, disinhibition and their interactions with gender as predictor variables, while the outcome variables were the six rotated components from the PCA analysis. Variables of no interest were BMI and age. Here, we did not find any significant relationships. Results of this analysis can be found in Tables S6 and S7 (n = 106).

| Aim 3: fMRI investigations in samples including participants with obesity-relationship of hemispheric bias and self-reported behaviours
Here, we investigated relationships between fMRI relative asymmetry indices (L − R)/(L + R) and approach/avoidance behaviours We found a significant interaction effect of BAS Drive and gender on the rotated component 6 (RC6 ; Table 4). This component is strongly influenced by the BA10 (Table 5). This suggests that in men increased left over right-brain activity in the BA10 is related to lower BAS drive scores, while in women increased left over right brain activity is related to higher BAS drive scores ( Figure 2). Furthermore, the results showed a significant interaction effect of BAS drive and gender on RC5, and a main effect of BAS drive on RC5 with contributions from the VTA (Tables 4 and 5). It suggests that in women increased left over right hemispheric activity in the VTA is related to increased BAS drive scores ( Figure 3). Finally, we also found a significant association between RC5 scores and BIS anxiety (Tables 4 and 5

| DISCUSSION
In this study, we aimed at replicating previous EEG findings concerning relationships of resting-state hemispheric asymmetries and approach/avoidance behaviours in healthy participants. Second, we aimed to investigate whether EEG asymmetry findings and fMRI asymmetry findings correspond to each other in the approach/avoidance context, as they do in the language (e.g. syntactic and semantic processing), or attention context (e.g. object or face perception; Chakrabarty et al., 2017;Mazza & Pagano, 2017;Powell et al., 2006).
Importantly, we also used fMRI to obtain data from subcortical structures, which are not easily obtainable from the EEG measures. This is an important addition especially in the context of obesity, since alterations in functions and structure of subcortical dopaminergic regions were previously often related to obesity (Cone et al., 2013;Friend et al., 2016;Geiger et al., 2009;Horstmann et al., 2015;Narayanaswami et al., 2013;Stice et al., 2011;Volkow et al., 2008;Vucetic et al., 2012). Furthermore, we attempted to expand the findings to self-reported eating behaviour and BMI (which has been related to increased approach behaviour; Mehl et al., 2019;Mehl et al., 2018) using rsfMRI. We tested three independent samples to answer these questions. In Sample 1, we were not able to directly replicate Sutton's and Davidson's EEG findings showing a positive association between BAS − BIS scores (describing individual differences between approach and avoidance behaviours) and higher left restingstate hemispheric bias. However, we show a conceptual replication of this bias with BAS drive in women. Second, we were not able to find significant associations between rsfMRI data and approach/avoidance behaviours in the same Sample. Furthermore, in Sample 2-which included participants with overweight and obesity as well as rsfMRI data-we found significant associations between hemispheric asymmetries, gender, BAS drive and BIS anxiety. Finally, in none of the samples did we find significant relationships of hemispheric bias and self-reported eating behaviour or BMI.
Past work by Grey and colleagues has suggested that human behaviour is driven by the interplay of the behavioural inhibition and activation systems (Gray, 1981;Gray & McNaughton, 1992). In a number of clinical and laboratory studies, it has been proposed that those fundamental behavioural dimensions are driven by asymmetric engagements of anterior brain regions (Davidson & Hugdahl, 1995;Harmon-Jones & Gable, 2018). In particular, the neural substrate for the inhibition system or withdrawal behaviour was found in the right prefrontal cortex, while the left prefrontal cortex was related to approach behaviour (Davidson & Hugdahl, 1995;Harmon-Jones & Gable, 2018). Those conclusions are based predominantly on rsEEG studies but also on studies in patients with frontal brain lesions. In our work, we aimed to replicate the seminal study by Sutton and Davidson (1997), which showed a positive association of BAS − BIS differential scores with left hemispheric bias, as measured by absolute alpha power from rsEEG. Although we have analysed our data in the same way, we did not replicate these results. In our study, the rsEEG Importantly, in a more detailed EEG data analysis using a refined relative asymmetry index, that is superior to an absolute in terms of interpretability, and relative alpha power, we found effects that are conceptually similar to the ones by Sutton and Davidson (1997) (Carver & White, 1994), but we used it nevertheless only to directly replicate findings of Sutton and Davidson (1997). It is possible that those different measures are related to hemispheric asymmetries in a distinct, gender-dependent way. Additionally, previous literature shows that gender indeed might influence hemispheric asymmetries-brains of men seem to be more lateralised as compared to women (Hausmann, 2002(Hausmann, , 2017McGlone, 1980). This does not exclude the possibility that women's brains show different associations between hemispheric asymmetries and self-reported behaviours, possibly through sex hormones (Hausmann, 2002(Hausmann, , 2017.
Future studies should aim to replicate our result and investigate asymmetries specifically with regard to gender differences.
It is worth noting that we found significant associations of questionnaire measures and hemispheric asymmetries measured with low relative alpha power, but not with broadband relative alpha power.
Since low alpha power represents such attentional processes as vigilance (Klimesch et al., 2007;Petsche et al., 1997), our results suggest that hemispheric asymmetries are related to those processes, rather than to general inhibitory processing within the brain.
The second aim of our study was to investigate whether approach/avoidance-related asymmetries can be measured with both EEG and fMRI. We were not able to replicate EEG findings in Sample 1 using rsfMRI. Such lack of replication might be related to the fact that alpha power and fALFF measure different processes. This is also reflected in a lack of direct relationship between EEG and whole-brain fALFF asymmetries. Alpha power is indeed conceptualised to be inversely related to brain activity by enabling active inhibition (Klimesch et al., 2007). fALFF, on the other hand, is generally suggested to be a measure of brain activity (Zou et al., 2008). For example, low-frequency fluctuations (LFFs) in grey matter were previously found to be higher than in white matter suggesting that they reflect grey matter metabolism and activity (Biswal, Yetkin, Haughton, & Hyde, 1995). This claim was further substantiated by a study which created a map of resting fluctuations in the visual cortex, suggesting that LFFs reflect spontaneous brain activity (Kiviniemi et al., 2000). Spontaneous LFFs were also identified in the default mode network at rest, again, suggesting that they might reflect brain activity (Fransson, 2005). We therefore hypothesised that alpha power and fALFF could simply be inversely related to each other. This is, however, not supported by our data. Instead, this relationship seems to be more complex. This might be because EEG and fMRI measure predominantly post-synaptic potentials and BOLD response, respectively (Bucci & Galderisi, 2011;Gauthier & Fan, 2019 We further investigated the relationship between hemispheric bias and BMI, since BMI in the obese range is related to increased approach behaviour (Mehl et al., 2018;Mehl et al., 2019) and obesity has been described as a deficiency of right-brain activation (Alonso-Alonso & Pascual-Leone, 2007). This was done in Samples 2 and 3, since they included participants with BMI in the overweight and obese range. Our analyses did not show a significant relationship between hemispheric bias and BMI. Thus, we did not find support for the right-brain theory of obesity, which suggests that hemispheric biases at rest may not be related to BMI per se, but to specific patterns of approach/avoidance and/or eating behaviour instead. Relatedly, it is conceivable that hemispheric biases during specific task performance might be related to BMI. While previous studies supporting the right brain theory of obesity largely focused on patients with unilateral brain lesions or structural asymmetries (Colcombe et al., 2006;Regard & Landis, 1997;Short et al., 2005;Uher & Treasure, 2005), our resting-state data were obtained in neurologically healthy participants. This may imply that previous results on obesity-related hemispheric asymmetries cannot be generalised to individuals with obesity. This heterogeneity, while increasing ecological validity, might introduce noise, which in turn makes it difficult to detect associations between BMI and hemispheric asymmetries. Finally, the right-brain theory of obesity is based on a number of findings relating eating behaviours and physical activity to hemispheric asymmetries, and not to BMI directly (Colcombe et al., 2006;Regard & Landis, 1997;Short et al., 2005;Uher & Treasure, 2005), as did our study-which might explain deviating results. In sum, future studies need to focus on relationships between obesity measures and hemispheric asymmetries in EEG and fMRI measurements of both resting-state and task contexts to confirm or revise the right-brain theory of obesity.
Finally, we investigated associations between hemispheric asymmetries and self-reported eating behaviours in all three samples.
Here, we did not find any relationships using rsEEG and rsfMRI data.
That is, we were not able to replicate previous rsEEG findings showing hemispheric bias relationships with disinhibition, hunger (Ochner et al., 2009) or restrained eating (Silva et al., 2002). Similarly, the study by Ochner et al. (2009) included participants with overweight and obesity (so did 2 of our 3 samples), and the study by Silva et al. (2002) included only lean women (one of our samples included mostly lean participants and we investigated interactions with gender). However, certain differences between those studies and our research exist, which might explain different results: First, Ochner and colleagues investigated a group of much older participants (mean age: 49 years).
It is conceivable that the duration of obesity influences prefrontal asymmetries, hence age might explain differences between results.
Furthermore, in our study, we were very conservative with regard to multiple comparisons correction, while Ochner and colleagues were more liberal in this respect.
Some limitations of our study need to be acknowledged: EEG data were only available for one sample. It would provide additional evidence to investigate differences between rsEEG and rsfMRI asymmetry associations with behavioural measures in other samples, especially concerning BMI and eating behaviour-aspects not investigated as thoroughly as approach/avoidance behaviours. As our study investigated relationships between self-reported approach/avoidance behaviours and resting-state neuroimaging measures, future studies could also include task-based neuroimaging measures, especially in the context of obesity. This might give a more valid proxy for everyday motivational behaviours and therefore have higher ecological validity.
In sum, we conceptually replicated findings showing relationships between hemispheric bias and approach/avoidance behaviours in women, but not self-reported eating behaviour in both rsEEG and rsfMRI. Moreover, we investigated relationships between rsEEG alpha power measures and rsfMRI fALFF. We show that associations of hemispheric asymmetries measured with rsEEG and rsfMRI are similar, however, we do not provide a replication of rsEEG results and rsfMRI results in the same sample. Future studies should answer the question of how those measures relate to each other in a more systematic way. We suggest that future studies should be performed using samples of lean, overweight and obese participants using both EEG and fMRI measures.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.