Executive functioning moderates neural reward processing in youth

Abstract

Although executive functioning has traditionally been studied in “cool” settings removed from emotional contexts, it is highly relevant in “hot” emotionally salient settings such as reward processing. Furthermore, brain structures related to “cool” executive functioning and “hot” reward-related processes develop simultaneously, yet little is known about how executive functioning modulates neural processes related to reward processing during adolescence, a period of time when these systems are still developing. The present study examined how performance on “cool” behavioral executive functioning measures moderates neural reward processing. Youths (N=43, mean age[SD]=13.74[1.81] years,) completed a child-friendly monetary incentive delay task during fMRI acquisition that captures neural responses to reward anticipation and to reward receipt and omission. Performance on inhibitory control and cognitive flexibility measures, captured outside the scanner, was used to predict brain activation and seed-based connectivity (ventral striatum and amygdala). Across analyses, we found that executive functioning moderated youths’ neural responses during both reward anticipation and performance feedback, predominantly with respect to amygdala connectivity with prefrontal/frontal and temporal structures, supporting previous theoretical models of brain development during adolescence. Overall, youths with worse executive functioning had more pronounced differences in neural activation and connectivity between task conditions compared to youths with better executive functioning. This study contributes to elucidating the relationship between “cool” and “hot” processes and our findings demonstrate that simple executive functioning skills moderate more complex processes that involve incorporation of numerous skills in an emotionally salient context, such as reward processing.

Introduction

Executive functioning is a set of skills related to the ability to voluntarily act in line with one’s internal goals (Crone et al., 2017; Luna, 2009; Miller et al., 2001; Miyake et al., 2012). Better executive functioning in youths is associated with many broad short- and long-term positive outcomes, including mental and physical health (Diamond, 2013; Moffitt et al., 2011; Wasserman et al., 2013), and may help youths adaptively navigate adolescence – a developmental period during which reward sensitivity (i.e., sensation- and novelty-seeking) is at its peak (Galvan, 2014; Silverman et al., 2015). Heightened reward sensitivity during adolescence promotes adaptive exploratory behaviors, which in turn facilitate cognitive development necessary for youths to eventually assume adult roles and behaviors (Romer et al., 2017; Silverman et al., 2015). Yet, risky and impulsive behaviors may also result from heightened reward sensitivity in adolescence and lead to neglecting threat-based cues (Van Leijenhorst et al., 2010). Furthermore, rates of depression (Thapar et al., 2012) and anxiety (Bandelow et al., 2015) also increase during adolescence and are linked to the same circuits implicated in heightened reward sensitivity (Benson et al., 2015; Chantiluke et al., 2012; Forbes et al., 2006). Importantly, better executive functioning has been shown to be protective against risky, impulsive behaviors (Luna, 2009; Miyake et al., 2012), as well as depression and anxiety symptoms (Han et al., 2016) in youths.

Prior work demonstrated that dysregulation of reward circuitry is related to increased reward sensitivity in adults with addiction disorders (Luijten et al., 2017) and has also shown that executive functions modulate similar reward processing systems in adults (Botvinick et al., 2015; Kouneiher et al., 2009; Yee et al., 2018). For example, preference for immediate smaller rewards over larger delayed rewards, an aspect of executive functioning, is related to increased ventral striatum activation in the context of receiving feedback regarding gain and loss of monetary reward (Hariri et al., 2006). Yet, little is known about how executive functioning modulates neural processes related to reward during adolescence when these systems are still developing. Understanding how this process unfolds in adolescence is important because it could play a role in identifying at-risk youths and help to elucidate how the interaction of these systems supports adaptive development. To this end, the present study aimed to investigate the moderating role of executive functioning on the neural correlates of reward processing.

Two core components of executive functioning which likely relate to reward processing are inhibitory control and cognitive flexibility. Inhibitory control is the ability to control one’s attention, behavior, thoughts, and emotions to override an internal drive or external pull; cognitive flexibility refers to mental set-shifting that allows for changing perspectives or approaches to a problem and flexible adjustment to new demands, rules, or priorities (Diamond, 2013). Importantly, although these two constructs have predominantly been studied separately, factor analytic work has demonstrated that they load onto a common executive functioning factor (Miyake et al., 2012). Executive functioning has traditionally been studied in “cool” contexts, i.e., using laboratory tasks that lack any obvious motivational significance, and without relating these skills to behaviors in “hot” emotionally salient contexts. Yet, these skills are highly relevant in such contexts, i.e., in situations that require reflective responding in the context of reward information or emotional salience (Diamond, 2013; Welsh et al., 2014). Behavior in “hot” contexts has been predominantly studied in terms of self-regulation, a construct that overlaps with executive functioning skills (Diamond, 2013; Jones et al., 2016) but has traditionally been studied separately from executive functioning research. Self-regulation research focuses on real-world behavior, for example, the ability to delay an immediate reward for a larger more desirable reward, and is typically assessed via adult ratings of children’s behavior in school and at home (i.e., real-world settings), or through observation of children in experimental settings that require them to delay gratification or endure a frustrating situation (Diamond, 2013). The ability to self-regulate in emotionally salient contexts may relate to basic subcomponents of self-regulation, i.e., executive functioning skills, such as inhibitory control and cognitive flexibility, measured in “cool” laboratory settings. For example, maintaining focus on a higher priority task (e.g., homework) when encountering a more immediately rewarding alternative task (e.g., checking social media) requires inhibitory control, or the ability to inhibit the urge for immediate rewards. Furthermore, adaptive responding to the disappointment of not receiving an expected reward (e.g., high exam grade) may require inhibiting an excessively frustrated reaction (inhibitory control) and cognitive reframing of the situation (cognitive flexibility). However, because “cool” and “hot” processes have traditionally been researched separately, it is not well understood how they relate to each other, especially whether basic executive functioning skills measured in a “cool” setting relate to how one responds in “hot” scenarios.

Brain systems involved in “cool” and “hot” processes work in tandem, yet the brain structures predominantly related to “cool” executive functioning processes and “hot” reward-related processes are often discussed separately in the literature and have been shown to develop concurrently but at different rates. A vast amount of evidence points to the prefrontal cortex as the brain structure that controls executive functioning (e.g., (Crone et al., 2017; Miller et al., 2001; Satterthwaite et al., 2013). It is thought that the prefrontal cortex allows for mental representations that provide “top-down” signals to other parts of the brain that can serve to regulate attention, and direct thoughts and behaviors in line with rules and internal goals (Miller et al., 2001). Furthermore, evidence suggests that top-down control of affective and non-affective tasks is related to the same prefrontal regions (Tabibnia et al., 2011). Maturation of the prefrontal cortex proceeds in a prolonged manner (Hsu et al., 2014) that extends until the early 20s (Cohen et al., 2016; C. F. Geier, 2013). In contrast, brain structures related to reward processing develop more quickly (Somerville et al., 2010) and are mediated by fronto-striatal circuitry (Galvan, 2014). Increased reward sensitivity during adolescence is thought to emerge because of the different rates at which the brain structures underlying the executive functioning and reward systems mature (Luna, 2009; Silverman et al., 2015; Somerville et al., 2010). For example, neuroimaging studies demonstrate age-related differences of striatum and amygdala activation in response to reward outcomes: adolescents vs. adults had greater striatal activation in response to favorable outcomes (Ernst et al., 2005; Van Leijenhorst et al., 2010), but less amygdala activation (Ernst et al., 2005). Additionally, adolescents vs. adults have less activation of frontal and parietal executive control regions during reward processing (Silverman et al., 2015). Indeed, theoretical models of neural network development in childhood (see (Shulman et al., 2016) for an overview) posit that the reward-driven socioemotional systems mature quickly compared to the more slowly maturing executive functioning system. More mature executive functioning networks may therefore dampen an overactive reward-system and thereby promote adaptive functioning. It has furthermore been suggested that the amygdala, which is linked to emotional salience, intensity, and avoidance, is also important for understanding the interaction between executive functioning and reward processing (Ernst, 2014; Richards et al., 2012). Indeed, individual differences in the emotional salience of cues likely play a role in how executive functioning and reward processing interact.

Empirical investigations of these theoretical models have thus far focused on examining neural networks while performing reward processing or cognitive tasks with and without incentives for good performance (e.g.,(Kray et al., 2018; Shulman et al., 2016). Evidence suggests that sensitivity to reward and punishment and difficulty adjusting to reward contingencies are mediated by neural systems involving the striatum and amygdala, as well as frontal regions. For example, in adolescents across a variety of reward-related contexts, there is increased activation in the ventral striatum when processing gain vs. loss, in the putamen and the amygdala in response to reward outcome vs. anticipation, and in the insula in response to anticipation vs. outcome (Silverman et al., 2015). There is also evidence that abnormalities in these networks may be related to internalizing and externalizing symptoms (Brotman et al., 2017; Dougherty et al., 2018; Harms et al., 2014). However, the degree to which basic executive functioning skills moderate these “hot” reward-related processes has not been examined. A better understanding of these processes would have important implications for addressing a current debate: Some have suggested that interventions aiming to improve executive functioning should be developed and implemented in real-world contexts to better translate to every-day functioning (Guare, 2014); however, others present evidence that before full integration of neural networks occurs (i.e., before adulthood), training in “cool” cognitive skills may have potential to translate into real world functioning (Murray et al., 2018). Yet, findings on the efficacy of cognitive training are mixed (Sala et al., 2019) and sparse with regard to transferring to “hot” contexts. It is thus unclear to what extent research and intervention work should focus on “cool” executive functioning skills, yet understanding how executive functioning and reward processing interact may inform future intervention and research directions. Characterizing the normative development of executive functioning in tandem with neural reward processing is thus crucial for informing future intervention work that aims to support youths in healthy development and in navigating through adolescence adaptively.

The present study aims to address this gap in the literature by examining how “cool” executive functioning behavioral performance moderates neural reward processing. Youths completed a child-friendly monetary incentive delay task during fMRI acquisition that captured neural responses to reward anticipation as well as reward receipt and omission. Performance on inhibitory control and cognitive flexibility measures, captured outside the scanner, was used to predict brain activation and seed-based connectivity (ventral striatum and amygdala). Ventral striatum and amygdala seeds were used in connectivity analyses based on the above outlined evidence on their involvement in reward processing and salience networks (Ernst, 2014; Galvan, 2014). Broadly, we expected that executive functioning would moderate reward-related neural processes and more specifically that brain activation in prefrontal and striatal networks, as well as, striatal and amygdala connectivity with the prefrontal cortex would be implicated. Building on prior research on the neural networks of reward processing during adolescence (Silverman et al., 2015) we furthermore expected ventral striatum and insula involvement during reward anticipation and amygdala and posterior cingulate involvement during reward feedback. We did not have specific condition-dependent hypotheses regarding the directionality of the relation between executive functioning and reward processing as previous findings highly depend on the specific reward contexts, task conditions, brain regions, and populations studied. Overall, however, based on evidence that executive functioning matures during adolescence (Luna, 2009), and that adults vs. adolescents demonstrate increased activation of frontal and parietal executive functioning networks during reward processing (Silverman et al., 2015), we expected that higher vs. lower executive functioning would be associated with increased frontal and parietal activation as well as increased limbic connectivity with those regions in the context of reward trials.

Method

Participants

Data from 43 youths (mean age=13.74 years, SD=1.81) were utilized. Families of these youths were originally recruited from the community to participate in one of three clinical trials (Subsamples 1–3) at local research clinics. The majority of the present sample (Subsample 1; n=34) was recruited from an ongoing clinical trial (NCT03176004) of an eight-week attention bias modification training program for anxiety. Youths were also recruited from a clinical trial (NCT01147614) testing the efficacy of brief behavioral therapy delivered in primary care (Subsample 2, n=14) and another clinical trial (NCT02021578) evaluating a Family Depression Prevention intervention program that targeted parents with a history of depression and their at-risk children (Subsample 3, n=4). Additional subsample details are provided in Supplement 1.

Subjects were excluded from the present study if they met clinal cut-offs in anxiety (score on the parent-rated Screen for Child Anxiety and Related Disorders (Birmaher, 1997) ≥25) or depression symptoms (score on parent-rated Mood and Feelings Questionnaire (Angold et al., 1987) ≥27) at the time of the scan. Eight subjects were excluded based on these criteria and one additional subject was excluded due to missing executive functioning data, leading to a final N of 43.

In addition to clinically significant anxiety or depression at the time of the scan, exclusion criteria included MRI contraindications (e.g., orthodontic braces) and presence of a major co-occurring neurological disorder. Although participants were recruited through three different sources, procedures were equivalent across recruitment sources. Data collection for the current study was conducted by the same staff using the same scanner between November 2016 and May 2018. As the entire sample was below 18 year of age, parental permission and child assent were obtained for all participants. The University of California San Diego Institutional Review Board, in joint agreement with the San Diego State University Institutional Review Board, approved all procedures. Participant characteristics are provided in Table 1. Participants received monetary compensation and a photo of their brain.

Table 1.

Demographic and clinical characteristics

Characteristic (N=43)
Gender (% female)	53.5%
Age (years)
Mean (SD)	13.74(1.81)
Range	9.69–17.51
Race, n (%)
White	22(51.2%)
African American	3(7.0%)
Multiracial	8(18.6%)
Other	5(11.6%)
Hispanic Ethnicity, n (%)	14(32.6%)
Executive Functioning Composite
Mean (SD)	101.22(7.36)
Range	75–112.5
Cognitive Flexibility
Mean (SD)	102.76(7.92)
Range	78–117
Inhibitory Control
Mean (SD)	99.68(7.99)
Range	72–112
Motion (in mm), post censoring
Mean (SD)	0.018(0.026)
Range	0.004–0.122

Note: SD=Standard Deviation; Cognitive

flexibility=uncorrected standard score on the NIH Toolbox Dimensional Change Card Sort task;

Inhibitory control=uncorrected standard score on the NIH Toolbox Flanker task. Missing data: race, n=6; ethnicity, n=4.

Executive Functioning

Two executive functioning (EF) tasks from the cognitive battery of the NIH Toolbox that measure cognitive flexibility and inhibitory control, respectively, were completed by youths on an iPad on the same day as the fMRI scan. These tasks have been validated for the use in children and adults, and show excellent reliability and convergent validity (Zelazo et al., 2013). A NIH-toolbox-based algorithm computed performance on each task based on accuracy and reaction time. As performance on the tasks was highly correlated (r=0.71) and because of evidence that cognitive flexibility and inhibitory control load on the same general executive functioning factor (Miyake et al., 2012), a composite executive functioning score was created by averaging the uncorrected standard scores representing performance on each task. This composite score was used for all analyses presented in the main text. To account for age-related performance differences in executive functioning (as well as developmental differences in neural functioning, as described in greater detail below), age was included as a covariate in all analyses.

Cognitive Flexibility.

The Dimensional Change Card Sort task measured cognitive flexibility (Beck et al., 2011). During this task, youths are presented with two target pictures that vary along two dimensions (shape [e.g., rabbit/boat] and color [e.g., white/green]). First, a word cue is presented which indicates whether cards should be matched according to shape or color. Then, the participant is shown a third shape that matches one of the other shapes. The participant is instructed to tap the correct shape based on the cue provided during each respective trial.

Inhibitory Control.

The Flanker task was administered to measure inhibitory control (Diamond, 2013). The participant is to indicate the left-right orientation of a centrally presented stimulus (an arrow pointing to the left or right) while inhibiting attention to four other arrows (two on each side of the target arrow) that may be pointing in the same (congruent) or the other direction (incongruent). The participant taps buttons on the bottom of the iPad screen to indicate the orientation of the central arrow.

Child-friendly Monetary Incentive Delay Task

While undergoing fMRI acquisition, participants performed a “Piñata task” (Dougherty et al., 2018; Helfinstein et al., 2013; Wiggins et al., 2017) during which they had the opportunity to “hit” a piñata via button-press. This task reliably captures neural correlates of reward processing during reward anticipation (learning whether the upcoming trial will be a full or empty piñata), and performance feedback (learning whether they successfully hit or missed the piñata) in children, reliably eliciting reward-related brain activation, including in the striatum, thalamus, insula, and prefrontal cortex (Helfinstein et al., 2013).

There were two types of trials during the task: reward trials during which participants had the opportunity to win a reward (i.e., four-star piñata), and no-reward trials during which they could not win any stars (i.e., empty piñata). Youths were instructed to “hit” all piñatas. Each trial began with an anticipation period in which the participant was cued as to whether they will have the opportunity to receive a reward during that round (2000ms), followed by a jittered delay period (2500–5500ms). Then, the participants were presented with a target (i.e., cartoon piñata) which they were instructed to “hit” by pressing a button to simulate striking the piñata. After the participants “hit” the piñata. a delay period followed (1500ms). If the participant pressed the button within the allotted time, the piñata breaks, indicating a hit; missed targets swing away (1500ms). There were four possible performance feedback scenarios: 1) reward/hit, 2) reward/miss, 3) no reward/hit, 4) no reward/miss. A basket was then shown either displaying stars (reward/hit condition) or empty (reward/miss or no reward conditions) (1500ms). Inter-trial intervals were jittered. Participants completed three runs, approximately 5 minutes each, with a total of 60 trials across all runs (30 reward, 30 no reward). Additional details are provided in Supplement 1 and previous work (Dougherty et al., 2018).

Neuroimaging Acquisition

A 3T General Electric MR750 Discovery MRI scanner with a Nova Medical 32-channel head coil was used to acquire anatomical and functional brain images. Multiband procedures allowed for improved inference of executive functioning and reward processing correlates. Additional details are outlined in Supplement 1.

fMRI Data Preprocessing

Preprocessing protocols were implemented using Analysis of Functional NeuroImages (AFNI; https://afni.nimh.nih.gov/afni/) and included slice-time correction, functional image realignment, EPI/anatomical registration, and non-linear registration to the Talairach template, followed by 4mm spatial smoothing and voxelwise scaling into units of percent signal change. Image volume pairs with frame-wise displacement >1mm were censored from individual level analysis. Mean frame-wise displacement (head motion) was <0.20mm across all participants and >90% of data remained per task condition following motion scrubbing.

fMRI Data Analysis

First-level models.

In addition to the regressors of interest outlined below, nuisance regressors were added across all first-level models to account for head motion in the x, y, z, roll, pitch, yaw directions and third-degree polynomials to model low-frequency drift.

Activation.

Individual-level general linear models were created during anticipation and feedback periods to provide estimates of brain activation. The regressor of interest during the anticipation period (Reward Condition [reward, no reward]) was convolved with AFNI’s ‘dmBLOCK’ basis function over a variable duration (4500–5500ms, depending on the length of the jitter). Regressors of interest during the feedback period (Reward Condition [reward, no reward] and Performance [hit, miss]) were convolved with the ‘BLOCK’ function over 4500ms. The analysis output consists of beta coefficients at each voxel for each condition (anticipation period: reward, no reward; feedback period: reward/hit, reward/miss, no reward/hit, and no reward/miss).

Connectivity.

Generalized psychophysiological interaction analysis (gPPI) (McLaren et al., 2012), using amygdala and ventral striatum (nucleus accumbens) seeds, calculated functional connectivity during reward anticipation and feedback, producing four connectivity analyses (one each for right and left nucleus accumbens, and right and left amygdala). These seeds were selected based on previous work on reward tasks (Dougherty et al., 2018; Ernst et al., 2005; Helfinstein et al., 2013), as well as their documented involvement in reward (Galvan, 2014) and salience (Ernst, 2014) networks. The Talairach atlas was used to identify seed regions in AFNI (left ventral striatum=136mm³; right ventral striatum=168mm³; left amygdala=1288mm³; right amygdala=1280mm³). Analyses resulted in voxel-wise images representing connectivity between each seed region and the rest of the brain for each condition.

Second-level models.

We utilized AFNI’s 3dMVM program to build whole brain ANCOVA models to evaluate associations between executive functioning (i.e., composite score) and reward-related neural activation and connectivity, separately for anticipation and feedback periods. Separate models evaluated whole-brain activation and connectivity for each of the four seeds. All models included age as a covariate to control for age-related differences in executive functions and neural functioning. Cluster thresholds for each analysis were calculated with AFNI’s 3dClustSim using the mixed-model spatial autocorrelation function (-acf) and the NN1 2-sided option, per the most recent recommendations on cluster correction (Cox, 2017). The cluster extent threshold differed slightly depending on analysis (ks≥63–65). We applied a conservative height threshold of p<.005. Clusters that emerged based on highest-level interactions were of primary interest (anticipation model: EF x Reward Condition; feedback model: EF x Reward Condition x Performance), but lower order interactions were also included.

In addition to the main analyses that utilized an executive functioning composite score, supplementary second-level analyses evaluated brain activation and connectivity in relation to cognitive flexibility and inhibitory control separately; the results of these analyses are presented in the Supplemental Materials.

Additional Analyses

To evaluate the impact of outliers as well as potential confounds (i.e., anxiety, depression, residual head motion, psychotropic medication use, and recruitment source), beta coefficients were averaged for each cluster and individual and then extracted to conduct additional analyses in SPSS. Clusters were also examined for gender interactions. Additionally, predicted brain values based on low (minimum) and high (maximum) scores for both inhibitory control and cognitive flexibility were calculated to aid in the interpretation of results and graphed for illustrative purposes.

Results

To summarize, across analyses we found that executive functioning moderated youths’ neural responses during both reward anticipation and performance feedback. That is, youths’ brain activation, and amygdala and ventral striatum connectivity, depended on their level of executive functioning. Table 2 lists clusters that emerged in contrasts of interest (i.e., contrasts that included executive functioning in the model); clusters that emerged in additional contrasts (e.g., task effects) are listed in Tables S3–6. Clusters that were outlier driven or situated fully in the hemispheric white matter are not discussed but are listed and identified in Table S2 (i.e., +outlier driven, ^White matter).

Table 2.

Significant clusters resulting in contrasts of interest of whole-brain analyses

Activation
ANTICIPATION PERIOD
Executive Functioning x Condition
k	F	x	y	z	BA	Region
80	16.1	3	−49	54	7	Right Precuneus
73	18.6	27	1	−30	38	Right Superior Temporal Gyrus
FEEDBACK PERIOD
Executive Functioning Main Effect
k	F	x	y	z	BA	Region
296	35.1	−55	−23	−24	20	Left Fusiform Gyrus, Left Inferior Temporal Gyrus
281	31.5	−9	23	60	6	Left Superior Frontal Gyrus
222	26.5	−11	−69	−2	18	Left Lingual Gyrus, Left Culmen, Left Declive
120	27.8	−3	−73	22	18	Left Cuneus, Left Precuneus
69	23.4	23	−69	−6	19, 18	Right Lingual Gyrus
Executive Functioning x Condition
k	F	x	y	z	BA	Region
198	28.4	−43	1	8	13	Left Insula, Left Inferior Frontal Gyrus
Executive Functioning x Condition x Performance
k	F	x	y	z	BA	Region
85	32.4	65	−19	−12	21	Right Middle Temporal Gyrus
84	20.2	−39	7	16	13	Left Insula
79	14.9	43	−51	40	40	Right Inferior Parietal Lobule
Connectivity: Left Ventral Striatum
ANTICIPATION PERIOD
Executive Functioning x Condition
k	F	x	y	z	BA	Region
542	20.2	25	−9	54	6	Right Middle Frontal Gyrus
130	23.2	5	−23	54	6	Right Medial Frontal Gyrus
93	14	−25	−65	24	31	Left Precuneus
FEEDBACK PERIOD
Executive Functioning x Condition
k	F	x	y	z	BA	Region
73	16.9	53	−61	8	39	Right Middle and Superior Temporal Gyrus
Executive Functioning x Condition x Performance
k	F	x	y	z	BA	Region
143	26.8	−7	−45	38	7, 31	Left Precuneus, Left Cingulate Gyrus
Connectivity: Right Ventral Striatum
FEEDBACK PERIOD
Executive Functioning Main Effect
k	F	x	y	z	BA	Region
164	25.3	23	59	4	10	Right Superior Frontal Gyrus, Right Middle Frontal Gyrus
66	19.4	11	−5	68	6	Right Superior Frontal Gyrus
Executive Functioning x Performance
k	F	x	y	z	BA	Region
191	22.7	19	61	4	10	Right Middle Frontal Gyrus, Right Superior Frontal Gyrus
86	19.4	57	−9	46	6, 4	Right Pre- & Postcentral Gyrus
Executive Functioning x Condition x Performance
k	F	x	y	z	BA	Region
1009	42.8	−19	−5	44	6	Left Precentral Gyrus, Left Medial Frontal Gyrus
136	18.5	−15	−49	52	7	Left Precuneus
117	22.6	9	−7	50	6	Right Medial Frontal Gyrus
107	21.7	−21	31	28	9	Left Medial Frontal Gyrus, Left Superior Frontal Gyrus
81	22.6	−55	−25	26	40	Left Inferior Parietal Lobule, Left Postcentral Gyrus
74	17.8	-25	-7	2	-	Left Putamen
Connectivity: Left Amygdala
FEEDBACK PERIOD
Executive Functioning x Condition
k	F	x	y	z	BA	Region
518	36	−25	−1	62	6	Left Superior Frontal Gyrus, Left Middle Frontal Gyrus
219	42.6	67	−33	8	22	Right Superior Temporal Gyrus
178	23	57	−59	4	20	Right Middle Temporal Gyrus, Right Inferior Temporal Gyrus
92	21.4	49	−63	−10	37	Right Fusiform Gyrus
84	18.2	−53	33	16	46	Left Middle Frontal Gyrus, Left Inferior Frontal Gyrus
72	16.3	21	1	4	−	Right Putamen
Executive Functioning x Performance
k	F	x	y	z	BA	Region
194	28.4	−49	29	−8	47	Left Inferior Frontal Gyrus
92	17.5	−39	29	22	46	Left Middle Frontal Gyrus
Executive Functioning x Condition x Performance
k	F	x	y	z	BA	Region
341	26.9	25	−91	−14	18	Right Cuneus, Right Lingual Gyrus
172	25.3	−29	−41	62	5, 7	Left Postcentral Gyrus, Left Superior Parietal Lobule
169	24.8	63	−31	6	22	Right Superior Temporal Gyrus, Right Middle Temporal Gyrus
125	24.3	−25	−3	62	6	Left Middle Frontal Gyrus, Left Superior Frontal Gyrus
115	21.9	−27	−91	−12	18	Left Inferior Occipital Gyrus, Left Fusiform Gyrus
106	25.7	−53	3	−4	21, 22	Left Superior Temporal Gyrus, Left Middle Temporal Gyrus
80	14.1	−7	−27	28	23	Left Cingulate Gyrus, Left Posterior Cingulate
Connectivity: Right Amygdala
ANTICIPATION PERIOD
Executive Functioning Main Effect
k	F	x	y	z	BA	Region
90	25	67	−23	−16	20	Right Inferior Temporal Gyrus
Executive Functioning x Condition
k	F	x	y	z	BA	Region
412	35.6	−49	43	−6	47	Left Inferior Frontal Gyrus
FEEDBACK PERIOD
Executive Functioning Main Effect
k	F	x	y	z	BA	Region
270	32	33	57	−4	10	Right Superior Frontal Gyrus, Right Middle Frontal Gyrus
87	24.3	−31	57	0	10	Left Superior Frontal Gyrus, Left Middle Frontal Gyrus
70	19.9	−1	−69	10	30	Left Cuneus, Left Posterior Cingulate
65	21.4	13	55	14	10	Right Medial Frontal Gyrus
Executive Functioning x Condition
k	F	x	y	z	BA	Region
141	33.6	13	25	24	24	Right Anterior Cingulate, Right Cingulate Gyrus
139	27.7	−43	3	4	13	Left Insula, Left Precentral Gyrus
128	21.1	49	7	2	22	Right Superior Temporal Gyrus
103	28.3	23	31	34	8	Right Middle Frontal Gyrus
80	21.3	7	45	−8	10	Right Medial Frontal Gyrus, Right Anterior Cingulate
76	18	11	37	6	24	Bilateral Anterior Cingulate
75	29.7	9	−53	8	30	Right Posterior Cingulate
Executive Functioning x Performance
k	F	x	y	z	BA	Region
217	27.9	−41	19	48	9, 8	Left Middle Frontal Gyrus, Left Superior Frontal Gyrus
161	29.7	−7	−35	34	31	Bilateral Cingulate Gyrus
153	29.5	−69	−29	−12	21	Left Middle Temporal Gyrus, Left Inferior Temporal Gyrus
117	34.5	19	−89	28	19	Right Cuneus
95	20.2	−35	17	−28	38	Left Superior Temporal Gyrus
85	33.4	65	−45	12	22	Right Superior Temporal Gyrus
76	20.8	−43	−81	28	39	Left Angular Gyrus, Left Middle Temporal Gyrus
66	28.7	23	−93	−10	18, 17	Right Inferior Occipital Gyrus, Right Fusiform Gyrus
Executive Functioning x Condition x Performance
k	F	x	y	z	BA	Region
808	27	−21	−17	54	6	Left Middle Frontal Gyrus, Left Precentral Gyrus
770	29.5	−45	23	22	9	Left Precentral Gyrus, Left Inferior Frontal Gyrus
641	33.2	37	−29	64	6	Right Medial Frontal Gyrus, Right Precentral Gyrus
246	17.6	25	−43	46	7	Right Precuneus, Right Superior Parietal Lobule
191	30.4	23	67	4	10	Right Middle Frontal Gyrus, Right Superior Frontal Gyrus
151	21.9	−23	−51	66	7	Left Postcentral Gyrus, Left Superior Parietal Lobule
143	22	−35	−43	36	7	Left Superior Parietal Lobule, Left Inferior Parietal Lobule
110	16.6	23	25	36	8	Right Middle Frontal Gyrus
89	23.2	−43	37	24	46	Left Middle Frontal Gyrus
75	25.5	13	−29	36	31	Right Cingulate Gyrus, Right Paracentral Lobule
72	23.5	−27	23	6	45	Left Anterior Insula

Note: BA=Brodmann area; Clusters significant at whole-brain-corrected threshold of p<.05 (see Method for details on cluster threshold); bolded clusters are presented in Figures 1 and 2; no significant clusters emerged in the analyses for any contrasts that are not listed. Additional cluster emerged in the analyses that were either outlier driven or situated in the white matter (>95% of voxels); these clusters are listed in Table S2 in the Supplement. For all contrasts df=41. Additional contrasts (e.g., task effects) are reported in the Supplement.

Reward Anticipation

During reward anticipation, executive functioning moderated brain activation in the right precuneus (Figure 1A) and right superior temporal gyrus. The direction of effects depended on the reward condition (i.e., Executive Functioning x Reward Condition). Youths with greater executive functioning showed greater right precuneus and right superior temporal gyrus activation when anticipating reward vs. when no reward was anticipated, whereas youths with lower executive functioning showed greater activation when no reward was anticipated vs. when a reward was anticipated (Figure 1A).

Figure 1. Executive functioning moderates reward-related brain activation.

Graphs display predicted brain activation values for indicated clusters based on best and worst executive functioning standard scores in the present sample (low=75, high=112.5), controlling for age. For corresponding scatterplots that do not account for age see Supplemental Figure S4. Brain images of additional clusters are displayed in Supplemental Figure S6. Brain images represent axial sections (left=left) with threshold set at whole-brain FDR-corrected p<.05.

Executive functioning also moderated right ventral striatum connectivity with the right middle frontal gyrus (Figure 2A), right medial frontal gyrus, and left precuneus, as well as right amygdala connectivity with the left inferior frontal gyrus. Here, too, the direction depended on the reward condition and the same pattern was observed across all clusters: compared to youths with better executive functioning performance, youths with worse performance had decreased connectivity during reward anticipation, but increased connectivity during no reward anticipation, whereas the opposite was observed among youths with better functioning. Youths with worse executive functioning had the opposite and a more pronounced connectivity pattern difference for reward vs. no reward conditions, compared to youths with better executive functioning (Figure 2A).

Figure 2. Executive functioning moderates reward-related brain connectivity.

Graphs display predicted brain connectivity values for indicated clusters based on best and worst executive functioning standard scores in the present sample (low=75, high=112.5), controlling for age. For corresponding scatterplots that do not account for age see Supplemental Figure S5. Brain images of additional clusters are displayed in Supplemental Figure S6. Brain images represent axial sections (left=left) with threshold set at whole-brain FDR-corrected p<.05.

One additional cluster demonstrated a significant relationship between executive functioning and right amygdala connectivity with the right inferior temporal gyrus regardless of reward condition (i.e., main effects of Executive Functioning); increased executive functioning was associated with increased connectivity across both anticipation conditions.

Performance Feedback

When youths received feedback on whether or not they had successfully broken the full or empty piñata (performance feedback), executive functioning moderated task-condition dependent brain activation (i.e., Executive Functioning x Reward Condition x Performance) in the left insula (Figure 1Bi), right middle temporal gyrus (Figure 1Bii), and right inferior parietal lobule. The effect was driven by brain activation differences in response to reward vs. no reward misses. For example, executive functioning moderated left insula activation such that youths with lower vs. higher executive functioning showed increased activation when an empty piñata was missed, but decreased activation when a full (reward) piñata was missed (Figure 1Bi). This pattern was opposite in the right middle temporal gyrus (Figure 1Bii) and the right inferior parietal lobule: youths with lower vs. higher executive functioning showed decreased activation when an empty (no reward) piñata was missed, but increased activation when a full piñata was missed.

Furthermore, the Executive Functioning x Reward Condition x Performance interaction revealed ventral striatum and amygdala connectivity with broad networks of frontal/prefrontal, parietal, temporal, occipital, as well as other limbic regions. For all right ventral striatum connectivity clusters, the effect was driven by connectivity differences in response to reward vs. no reward hits as well as misses. For example, youths with lower vs. higher executive functioning showed decreased right ventral striatum connectivity with the left putamen when an empty piñata was hit (no reward), but increased connectivity when a full piñata (reward) was hit; they also showed increased connectivity when an empty (no reward) piñata was missed, but decreased connectivity when a full (reward) piñata was missed (Figure 2Bi). For all other connectivity clusters in the Executive Functioning x Reward Condition x Performance contrast (i.e., left ventral striatum, and right and left amygdala seeds) the effect was driven by connectivity differences in response to reward vs. no reward misses. For example, youths with lower vs. higher inhibitory control showed decreased right amygdala connectivity with the left insula when an empty (no reward) piñata was missed, but increased activation when a full piñata was missed (Figure 2Bii).

Activation and connectivity analyses examining the performance feedback period additionally revealed significant clusters for the main effect of Executive Functioning as well as lower level interactions (i.e., Executive Functioning x Reward Condition, Executive Functioning x Performance) across broad brain networks (see Table 2 for details). Overall, during performance feedback, across all significant interaction clusters in all contrasts, compared to youths with higher executive functioning performance, youths with lower executive functioning evinced the opposite and more pronounced reward vs. no reward differences in brain activation and connectivity.

Additional Analyses

Additional analyses indicated that results were not primarily driven by gender, anxiety, depression, head motion, psychotropic medication use, or recruitment source. We also evaluated whether the results differed by gender by allowing gender to interact with the effect of interest in each cluster. None of the reported findings differed by gender. Clusters that were outlier driven are not presented as part of the main findings but are listed and identified in Table 2. See Supplemental Materials for details.

Discussion

To our knowledge, this is the first study to demonstrate that youths’ executive functioning performance in “cool” contexts modulates neural responses during reward processing, that is, in a “hot” context. In line with our hypotheses, we found that executive functioning moderates children’s neural responses when they anticipate rewards and when they receive feedback on whether they obtained or missed a reward. As such, the present study bridges research in self-regulation that focuses on complex cognitive and behavioral processes in the context of emotionally salient situations, such as reward processing, and research in cognitive science that has traditionally studied executive functioning removed from such contexts.

In line with our hypotheses, broad neural networks involving predominantly prefrontal/frontal, parietal, and limbic areas were implicated. These findings align well with work outlining the neural networks involved in cognitive control and reward processing mechanisms (Casey et al., 2019; Shulman et al., 2016; Silverman et al., 2015; Somerville et al., 2010). Furthermore, we found that the neural networks that were affected by executive functioning performance differed depending on reward phase. For example, in line with previous literature (Silverman et al., 2015), ventral striatum connectivity was more broadly implicated during reward anticipation compared to amygdala connectivity, which in turn was more predominantly relevant during performance feedback. This suggests that when youths encounter a potential reward (reward anticipation), executive functioning skills may help modulate the perceived values of rewards (mediated by striatal networks). Furthermore, executive functioning skills may help youth modulate the perceived salience of a reward and their affective reaction (mediated by amygdala networks) when they learn whether they did or did not receive a reward (reward receipt/omission). Importantly, our findings related to amygdala connectivity provide support for theoretical ‘triadic’ models that have proposed that the amygdala is important to consider in models of neural development during adolescence (Ernst, 2014), in addition to the striatum and prefrontal cortex.

Overall, we found that youths with worse executive functioning had more exaggerated neural responses (i.e., greater differences in neural activation among conditions) – and such differences were in the opposite direction – compared to youths with better executive functioning, when anticipating and receiving feedback about a reward. This may indicate that among youth with worse executive functioning, neural networks are less mature or integrated, and thus less similar to those of adults. In turn, better executive functioning may be protective in that it may help youths with balanced and restrained responses in the context of reward processing that are more similar to those of adults (Luna, 2009; Silverman et al., 2015). Indeed, similar to previous findings comparing adolescents vs. adults (Silverman et al., 2015), higher executive functioning was associated with increased activation in parietal areas during reward vs. no reward anticipation, as well as increased limbic connectivity with frontal areas. Our findings also align with the literature on executive functioning in pediatric psychopathology that demonstrate worse executive functioning among youth with internalizing (Han et al., 2016) and externalizing symptoms (Hobson et al., 2011). For example, better inhibitory control is related to more effective regulation of anger in children (Cole et al., 2011; Gagne et al., 2011), and better cognitive flexibility predicts better coping in the context of depressive symptoms (Evans et al., 2016). The present study suggests that better executive functioning may protect youths from exaggerated neural responses in the context of emotionally salient processes and, as an extension of these findings, it may also enable youths who face emotional challenges to engage neural circuits that promote adaptive behavior and mental health.

The present study, in combination with prior literature, also suggests that executive functioning and reward processes may influence each other in a bidirectional fashion. Previous research demonstrated that rewards affect executive functioning performance, that is, incentives improve youths’ performance on executive functioning tasks. For example, although adolescents typically perform worse on executive functioning tasks compared to adults, their performance is comparable to adults when they are rewarded for good performance (C. Geier et al., 2009; Qu et al., 2013), although there remain differences in neural functioning between youths and adults while engaged in the executive functioning task (C. F. Geier et al., 2010). The present study provides complementary information because it demonstrates that basic non-incentivized executive functioning skills impact neural reward processing. Taken together, previous research and the present study suggest bidirectionality in reward and executive functioning processes, although studies specifically designed to test bidirectionality will be needed to confirm this.

Finally, our findings that basic executive functioning skills moderate reward processing are relevant to an ongoing debate over whether training in basic executive functioning skills may translate to changes in other contexts, such as situations in which a potential reward is present. Extant evidence on computerized cognitive training in healthy adults suggest that performance gains do not transfer to improvements in everyday functioning (Kable et al., 2017), yet evidence is mixed in children (Dovis et al., 2015; Sala et al., 2019). For example, cognitive training in basic attention has been shown to improve children’s ability to delay gratification (Murray et al., 2018), a “hot” emotionally salient process. Developmental “hypo-frontality” (i.e., prolonged maturation, and thus plasticity, of the prefrontal cortex) (Hsu et al., 2014) may be an evolutionary advantage that could be leveraged through executive functioning training. Among youths, whose prefrontal cortex is not fully developed, such training may more easily transfer to other skill domains, whereas in fully developed brains that are optimized for rule-based performance, transfer of such skills is more difficult (Hsu et al., 2014). Youths’ “hypo-frontality” may therefore be beneficial and cognitive training interventions could be helpful in reducing cognitive gaps in children (Hsu et al., 2014). The present study demonstrates that executive functioning skills based on inhibitory control and cognitive flexibility performance are implicated in reward processing and may inform the expansion of integrative prevention and intervention efforts.

Limitations

We note several limitations. First, our sample size (N=43) was modest, yet greater than many other studies examining the neural mechanisms of reward processing in adolescents (e.g., Ns=26 (Jarcho et al., 2012), 24 (Bjork et al., 2010), 16 (Ernst et al., 2005)). Second, participants for the present study were recruited from multiple clinical trials that targeted various psychopathological phenotypes. Although participants with clinical levels of anxiety and/or depression were excluded from this study, it is possible that the nature of the samples impacted the results. Consequently, replicating our findings in a larger sample with more comprehensive screening for history of psychopathology will be necessary. Third, the cross-sectional design of our study does not allow for inferences regarding causality. Longitudinal studies beginning much earlier in development are needed to characterize these processes as the neural circuitry associated with reward processing and cognitive control matures and to delineate how these neural circuits relate to behavioral outcomes. Finally, we chose to prioritize cognitive flexibility and inhibitory control based on previous literature implicating that these executive functioning processes are involved in reward processing (Brotman et al., 2017), and thus, our measures of executive functioning are not comprehensive. Multiple, more comprehensive measures that include other subcomponents, such as working memory, would allow a broader examination of executive functioning effects on reward processing as well as factor analytic approaches to assess common executive functioning factors vs. unique contribution of the individual components (Miyake et al., 2012).

Conclusion and Future Directions

The present study contributes to a growing body of literature aiming to elucidate the relationship between “cool” and “hot” processes during adolescence and how these important mechanisms (i.e., executive functioning and reward processing) interact on a neural level. Our findings demonstrate that basic “cool” executive functioning skills moderate more complex processes that involve incorporation of numerous skills in an emotionally salient reward processing context. We investigated top-down executive functioning on bottom-up reward processing, yet a number of studies have also demonstrated that reward incentives improve performance on executive functioning tasks (C. Geier et al., 2009; Qu et al., 2013). It will therefore be important for future work to elucidate the bidirectional nature of the interactions between executive functioning and reward circuits. Furthermore, it is important to point out that the current study used laboratory tasks to probe both “cool” and “hot” processes, and it is unclear whether the neural processes observed during the piñata task capture the complexity of “hot” scenarios in real-life which include immediate tangible rewards and consequences, environmental influences, and the presence of peers and/or adults. Further research is therefore needed to investigate to what extent the present findings translate to “hot” scenarios that involve such real-life complexities. As adolescence is a period of time during which reward sensitivity is at its peak and risk for psychopathology increases, understanding these mechanisms may help inform prevention and intervention development to help youths at risk navigate adolescence adaptively. Investigating the development of executive functioning and reward circuits, and how environmental and emotional factors influence these longitudinal changes are therefore important directions for future research. For example, in addition to cognitive training, investigating the role of physical exercise, social connection, positive emotions, and mindfulness in supporting adaptive brain development may be fruitful avenues for further exploration.