Conceptualizing Avoidant/Restrictive Food Intake Disorder via an Executive Functioning Lens

Abstract

Objective:

Avoidant/restrictive food intake disorder (ARFID) is a heterogeneous disorder wherein restrictive eating is primarily attributed to non-shape/weight-based reasons (e.g., sensory sensitivity) that empirical research continues to explore. Mounting evidence suggests that ARFID often presents alongside neurodevelopmental diagnoses (NDs) or divergent neurodevelopment broadly. Executive functioning (EF) differences often characterize divergent neurodevelopmental trajectories. Additionally, restrictive eating in anorexia nervosa has been conceptualized as related to EF factors (e.g., set shifting). Given the neurodevelopmental phenotype that may be associated with ARFID and the role of EF in anorexia nervosa, this paper proposes EF as a potentially important, yet understudied factor in ARFID pathology.

Method:

We posit that various observed ARFID behavioral/cognitive tendencies can be conceptualized in relation to EF differences. We contextualize commonly observed ARFID presentations within ‘core’ EF components (i.e., cognitive flexibility, working memory, inhibitory control), leading to hypotheses about EF in ARFID. Finally, we offer additional considerations/directions for future research on EF in ARFID.

Results:

Increased research on EF in ARFID is needed to consider this potential common factor in the etiology and maintenance of this heterogeneous disorder.

Discussion:

We aim to promote further consideration of EF in ARFID etiology, maintenance, and treatment-outcome research.

Introduction

Avoidant/restrictive food intake disorder (ARFID) is a heterogeneous disorder characterized by restrictive eating without weight/shape dissatisfaction (American Psychiatric Association, 2013). ARFID is associated with neurodevelopmental diagnoses (NDs) — autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), intellectual developmental disorder, and learning disorders (Bourne et al., 2022; Farag et al., 2022; Reilly et al., 2019; Stone-Heaberlin et al., 2020; Nicely et al., 2014) — and related ‘traits’ (e.g., attentional, social, and communication difficulties) (Inoue et al., 2021; Dinkler et al., 2022). Aspects of neurodevelopment like executive functioning (EF) thus may pertain to ARFID.

EF encompasses neurocognitive processes— cognitive flexibility, working memory, and inhibition— that engender goal-directed thought/behavior (Baggetta & Alexander, 2016; Suchy, 2009; Diamond, 2013). EF differences characterize NDs (Otterman et al., 2019). Moreover, ARFID shares features (e.g., eating-related rigidity) with anorexia nervosa (AN), wherein EF differences mark illness (Stedal et al., 2021) and recovery (Tchanturia et al., 2004; Lindner et al., 2014).

Given relatively increased neurodevelopmental differences in ARFID (Sanchez-Cerezo et al., 2023) and inflexibility that characterizes restrictive eating (Rodgers et al., 2023), we propose EF is understudied in ARFID. Using knowledge about ARFID and relevant presentations (e.g., other EDs, anxiety), we consider EF’s relationship to ARFID.

EF and ARFID Pathology

Diagnostic criteria suggest non-exhaustive, non-mutually exclusive examples (sensory sensitivity, limited interest in eating/food, feared consequences of eating) of why restrictive eating occurs in ARFID that frequently present in research (Reilly et al., 2019; Norris et al., 2018; Zickgraf et al., 2019; Sanchez-Cerezo et al., 2024; Katzman et al., 2022). Additionally, ‘preference for sameness’ has been a proposed presentation (Bourne et al., 2020). We review relevant research, outline EF components’ potential relationship to ARFID, and provide modeling considerations.

Cognitive Flexibility

Cognitive flexibility permits shifting perspectives and adjusting given changing circumstances (Diamond, 2013). Clinical observations suggest inflexible eating-related thoughts/behaviors characterize ARFID (Thomas & Eddy, 2019; Menzel et al., 2019), yet few studies test this. Reviewing research on inflexibility/ED risk and on inflexibility in ARFID-type pathology leads to our hypotheses.

Inflexibility-related constructs may predict restrictive eating, though this literature (versus corresponding cross-sectional literature) is small. In population-based studies, early-childhood cognitive inflexibility predicted ED symptoms later in development (Steegers et al., 2021; Dufour et al., 2023). Another related factor in non-ARFID EDs is poor central coherence (Lang et al., 2014), wherein ‘global processing’ impairments (interpreting ‘wholes’) and increased ‘local processing’ (interpreting ‘parts’) occur. Research suggesting both constructs are restrictive eating familial-risk factors provides support despite lacking longitudinal research (Lang et al., 2016; Kanakam et al., 2013).

Some research characterizes inflexibility in ARFID versus AN, citing shared eating-related inflexibility and detail-oriented processing as rationale (Mahr et al., 2023). This study indicated that, relative to child/adolescent AN, child/adolescent patients with ARFID demonstrated poorer switching, sequencing, and visual-attention alternation unexplained by weight-status differences on a set-shifting task, whereas Basile et al. (2021) identified no ARFID/AN set-shifting differences. Cross-sectional studies across ages suggest inflexibility and selective eating are related (Zickgraf et al., 2022; Foinant et al., 2022), altogether suggesting inflexibility in ARFID warrants study.

One hypothesis is that inflexibility facilitates a trajectory wherein low openness to dietary-variety expansion maintains childhood neophobia, as also suggested by other authors (Thomas & Eddy, 2019; Dovey et al., 2019). Inflexibility may promote/maintain predictions about undesirable outcomes (e.g., feeling disgusted, uncomfortable/feared physical sensations), problems integrating fear-incongruent evidence, and concrete food-related categorization. Additionally, reduced central coherence may underpin certain characteristics (e.g., over-attending to food characteristics/physical sensations). In summary, we hypothesize that inflexibility and related constructs (i.e., central coherence) may contribute to ARFID— particularly selectivity, rigid eating-related cognitions, and eating-related stimuli processing.

Working Memory (WM)

WM permits mental maintenance and use of previously presented information/stimuli (Diamond, 2013). Reville et al. (2016) proposed WM difficulties exacerbate decision-making difficulties in AN. Given similar cognitive styles in ARFID and AN, WM may bear consideration in ARFID. We review research on WM-related restrictive-eating risk, anxiety-related WM dysfunction and reduced WM, and WM in ARFID-type pathology, leading to hypotheses.

Some longitudinal research implicates WM in restrictive-eating onset. Schaumberg et al. (2020) found support for reduced childhood WM (i.e., decreased holding/use of information in WM) predicting restriction; authors proposed reduced WM manifests in dietary-planning difficulties that promote restriction. Superior WM predicted eating in response to positive external cues. Childhood WM predicted ED symptoms via cognitive inflexibility in Dufour et al. (2023). Theoretical work on memory in EDs broadly drew parallels to WM in anxiety (Forester et al., 2023), suggesting impaired exclusion of disorder-salient information from WM similarly characterizes EDs. Overrepresentation of disorder-relevant cues in WM occurs across anxiety research (Reinecke et al., 2008; Stout et al., 2013; Gambarota & Sessa, 2019). Non-clinical samples demonstrate associations between reduced WM and decreased fear extinction/discrimination (Stout et al., 2018; Laing et al., 2019).

No studies test WM in ARFID versus healthy controls, and Mahr et al. (2023) reported no ARFID/AN differences. WM was not associated with food rejection in Foinant et al.’s (2022) general-child sample, whereas in Dufour et al. (2023), reduced WM corresponded to picky eating. Evidence is mixed, with some indication reduced WM relates to ARFID-type pathology.

Based on previously cited anxiety-focused literature and theory regarding WM in EDs, we hypothesize that, in ARFID presentations with prominent fear/dislike of tastes/other sensations, WM may be biased toward disorder-relevant cues (e.g., food characteristics, physical sensations) despite absent threat. Reduced WM may impair fear-related learning, promoting ARFID maintenance. Reduced WM also predicts lower eating in response to positive cues; reduced WM may increase susceptibility to negative eating-related cues in ARFID, given hypothetically lower positive-cue responsivity. Conversely, low-appetite presentations may demonstrate dysfunctional processes that hold food preferentially in WM in healthy controls (Rutters et al., 2015). Finally, reduced WM may promote restriction via related neurocognitive difficulties (e.g., planning, cognitive inflexibility). In summary, we hypothesize WM dysfunction may translate to overrepresentation or underrepresentation of eating-related stimuli in WM (depending on ARFID presentation), impaired fear-related learning, and restricting via interplay between WM and related functions.

Inhibitory Control (IC)

IC allows for the direction of behavior/cognition to meet situational demands (Diamond, 2013). Conceptualizing IC as a spectrum, AN typically corresponds to ‘high’ IC (Brooks et al., 2012), whereas ARFID has not been situated within this framework. We review research on IC-related constructs in ARFID and on IC in anxiety, leading to hypotheses.

Basile et al. (2021) identified reduced IC in ARFID relative to AN. Related work on delay discounting (i.e., devaluing temporally distant rewards) demonstrated that ARFID corresponded to more immediate-gratification tendency relative to AN (Stern et al., 2024). One interpretation was that safe-/preferred-food reliance exemplifies immediate-reward preference, whereas ‘reward’ for changing preferences is distal (requiring effort/negative-emotion tolerance). Therefore, reduced IC may promote difficulties inhibiting food preferences and preference-maintaining factors (e.g., avoidance, over-attendance to disorder-relevant stimuli).

Within a cognitive-behavioral ARFID model, avoidance occurs across presentations (Thomas & Eddy, 2019; Thomas et al., 2018), like anxiety disorders. Attentional Control Theory (Eysenck et al., 2007) proposes anxiety is characterized by processing of disorder-salient stimuli over goal-directed processing (i.e., poor IC). IC and anxiety interact, with reduced IC precipitating anxiety and anxiety weakening IC (Shi et al., 2019; Roxburgh et al., 2019); bidirectionality may also occur in ARFID.

Thus, based on related preliminary work in ARFID (Basile et al., 2021; Stern et al., 2024) and conceptualizations of IC in anxiety (Eysenck et al., 2007), we hypothesize that reduced IC may contribute to difficulty inhibiting avoidance and redirecting attention when feared internal (e.g., nausea, fullness, disgust) or external (e.g., illness in one’s environment, unappealing food stimuli) cues arise. Likewise, eating-related anxiety may impair inhibition of cue-related attention and avoidance.

Additional Considerations

As noted, NDs occur in an ARFID sub-group; one-in-four patients may have ASD (Norris et al., 2018) and two-in-five patients may have ADHD (Reilly et al., 2019). NDs may moderate proposed relationships between EF and ARFID, such that relationships are stronger when ARFID and NDs co-occur, given EF differences in NDs (Otterman et al., 2019). EF could mediate relationships between ND characteristics and ARFID pathology and/or display differential relevance depending on present NDs. We do not propose NDs change the nature of hypothesized relationships between EF/ARFID, but that NDs necessitate modeling (Figure 1). Relatedly, parsing influences of other ARFID comorbidities that impact EF like depression (Bredemeier et al., 2016) and anxiety (Zainal & Newman, 2018) may be needed.

Figure 1. Incorporating Executive Functioning Into a Model of ARFID.

Note. This paper primarily focuses on proposing the possible executive functioning differences to restrictive eating/avoidance behavior pathway. Primary ARFID maintaining factors (e.g., proposed examples in current diagnostic criteria [American Psychiatric Association, 2013] and neurobiological conceptualization of ARFID [Thomas et al., 2017] such as sensory sensitivity and eating-related fears) are also depicted as contributors to restrictive eating/avoidance via the red arrow from “Possible Primary ARFID Maintaining Factors” to “Restrictive Eating and Avoidance Behaviors”. As depicted and as briefly described in our paper, given the positive relationship between diet quality and executive functioning, ARFID behaviors may also contribute to executive functioning difficulties in ARFID (e.g., lower diet quality affecting executive function in an acute manner and longer term). Neurodevelopmental diagnoses (or a broader phenotype) may promote executive functioning differences in ARFID (arrow from “Neurodevelopmental Diagnoses/Phenotype” to “Possible Executive Functioning Differences”) and thus contribute to “Restrictive Eating and Avoidance Behaviors,” or may moderate proposed relationships between executive functioning and ARFID behaviors (arrow from “Neurodevelopmental Diagnoses/Phenotype to the pathway between “Possible Executive Functioning Differences” and “Restrictive Eating and Avoidance Behaviors”). “Neurodevelopmental Diagnoses/Phenotype” may contribute to certain “Possible Primary ARFID Maintaining Factors” due to factors such as sensory sensitivity often occurring in such neurodevelopmental presentations. The relationship between primary maintenance mechanisms and ARFID behaviors may be moderated by executive functioning differences (as described in ‘Research Directions’). Finally, it is possible that the relationship between executive functioning differences and primary ARFID maintaining factors is bidirectional in certain cases, such that executive functioning differences (as described throughout our paper) may further solidify certain maintaining factors (e.g., working memory biases promoting attention to fears, cognitive inflexibility promoting attention to certain sensory characteristics of foods) and certain primary maintaining factors may also exacerbate executive functioning differences (e.g., eating-related fears reducing working memory and inhibitory control in eating-related situations).

Although we focus on proposing EF contributions to ARFID, dietary patterns in ARFID could influence EF, given nutrition/diet and early-feeding difficulties may influence neurodevelopment and increase EF difficulties (Costello et al., 2021; Koca & Huri, 2022). Diet quality— which is often decreased in ARFID (Harshman et al., 2019; Middleman et al., 2021)— promotes EF (Cohen et al., 2016). Longitudinal, population-based studies suggest fruit-/vegetable-rich diets in infancy/early childhood promote improved later EF (Nyaradi et al., 2013; Gale et al., 2009). Given ARFID often onsets in childhood (Kurz et al., 2015) and causes reduced fruit/vegetable consumption (Schmidt et al., 2021), ARFID could detrimentally affect EF and, in lifelong/childhood-onset presentations, its development. Given research on diet quality/EF in non-ARFID samples, studying EF in ARFID likely necessitates modeling the effect of diet on EF (Figure 1) to identify EF difficulties as inherent versus secondary to diet.

Finally, ARFID is currently diagnosable across the lifespan, necessitating consideration of developmental variability in EF’s role. EF refines over childhood/adolescence (Hendry et al., 2016; Best & Miller, 2010). Given EF is often assessed relative to age norms, relative EF differences are worth studying in ARFID across ages, particularly as cognitive functions promote feeding-skill development (Goday et al., 2019). Table 1 presents examples of tasks that assess aspects of EF as early as infancy. However, EF differences may most reliably correspond to ARFID in adolescents/adults, when EF increasingly matures/stabilizes (Anderson, 2002; Friedman et al., 2016).

Table 1.

Broad Hypothesized Roles of Executive Functioning in ARFID and Means of Assessing EF

Cognitive Flexibility	Working Memory (WM)	Inhibitory Control (IC)
Definition: ability to adjust cognitions (and in turn, behavior) to adapt to changing environmental demands	Definition: ability to cognitively maintain/use mentally represented information	Definition: ability to purposefully direct cognitive resources and behavior adaptively rather than respond impulsively or habitually
•Hypothesized role: may maintain childhood neophobia that persists beyond a developmentally typical degree by limiting openness to dietary expansion •Hypothesized role: may promote inflexible cognitions about food and eating-related outcomes	•Hypothesized role: WM dysfunction affects the extent to which eating-related cues are represented in WM •Hypothesized role: Reduced WM impairs fear-related learning related to eating-related outcomes •Hypothesized role: may interact with other functions (e.g., inflexibility, planning) to maintain restriction	•Hypothesized role: Reduced IC may complicate the inhibition of attention to disorder-salient cues (e.g., sensory aspects of food, uncomfortable physical sensations) and/or the inhibition of safety/avoidance behaviors
Assessment Examples •looking/reaching ‘A-not-B’ tasks adapted for infants (e.g., Bell & Adams, 1999), card sorting tests (e.g., Doebel & Zelazo, 2015), reversal learning tasks (e.g., Minto de Sousa et al., 2015)	Assessment Examples •visual paired comparison tasks for infants (e.g., Richmond et al., 2007), list sorting tests (e.g., Tulsky et al., 2013), span or updating tasks (e.g., Morra et al., 2021; Özdemir & Ganea, 2020; Schmiedek et al., 2009)	Assessment Examples •’Early Childhood Inhibitory Touchscreen Task’ for infants (e.g., Holmboe et al., 2021; Lui et al., 2021), Flanker tasks (e.g., McDermott et al., 2007), Go/No-go tasks (e.g., Zinchenko et al., 2019), Stop-signal tasks (e.g., Carver et al., 2001)

Note. This table organizes and summarizes hypothesized ways that core aspects of executive functioning (defined in the table) may be related to various manifestations of ARFID pathology. We also note non-exhaustive examples of how these aspects of executive functioning may be assessed, including in infancy/early childhood given that ARFID can present very early on in development. Tasks noted as being appropriate for infants include specific versions of tasks that are considered developmentally appropriate for infants, whereas remaining tasks are commonly used in varying forms from childhood through adulthood. Readers interested in more comprehensive information about assessing executive function in children as young as age 3 can reference the National Institutes of Health Toolbox – Cognition Battery (Zelazo et al., 2013).

Discussion

Research Directions

We provided testable hypotheses for future research on core-EF components in ARFID and additional factors to consider: comorbid NDs, diet/feeding difficulties, and EF developmental stability (Table 1/Figure 1). Employing varied EF measures would permit exploration of core and related EF manifestations (e.g., task initiation, attention).

Studies should assess NDs to ascertain the extent to which ND covariates account for EF and if EF is explained by NDs versus a broader phenotype including EF differences; ARFID may share genetic origins with NDs, with genetic propensity manifesting in co-occurring ARFID/threshold NDs or ARFID with ND characteristics.

Studies should clarify if EF difficulties result from active ARFID (e.g., via low weight and/or diet quality), contributed to ARFID onset and/or exacerbation, or both, necessitating longitudinal research on unidirectional versus bidirectional pathways. Longitudinal research is also needed to clarify the degree to which EF differences and development are accounted for by aspects of ARFID that can influence EF (e.g., diet quality). As ARFID occurs across weight statuses, EF and weight status interplay also necessitates consideration, given evidence for differential reward-related neurocognition in ARFID by weight status (Kerem et al., 2022).

Examining developmental differences in the role of EF in ARFID warrants research across ages. Studies might consider EF’s interactions with course and primary-maintenance mechanisms; WM bias may be notable in acute-onset, fear-based presentations, wherein hypervigilance to feared cues follows a frightening eating-/illness-related experience (Brigham et al., 2018; Zickgraf, 2018), whereas inflexibility may be notable in selectivity-based, lifelong presentations. EF may moderate relationships between primary-maintenance mechanisms and ARFID behaviors (Figure 1; e.g., sensory sensitivity more strongly predicting restriction/avoidance given increased inflexibility, fear more strongly predicting restriction/avoidance given increased WM dysfunction), partially mediate primary-maintenance mechanisms’ contributions to behaviors (Figure 1), or represent an additional ARFID phenotype.

Finally, it is unclear for whom ARFID interventions perform differentially (Willmott et al., 2024). If EF is a putative factor (or relevant for patient subsets), research might consider whether EF-related treatment alterations (Richson & Deville et al., 2024) improve outcomes. In cognitive-behavioral anxiety interventions (which share similarities with ARFID interventions), EF change predicts outcomes (Godovich et al., 2020), suggesting EF could be a target to empirically test. For example, flexibility promotes integrating exposure learning while IC prevents avoidance and maximizes attention (Craske et al., 2014), suggesting certain EF differences may impede exposure effectiveness.

Conclusion

Considering EF in ARFID could yield more comprehensive models for this heterogeneous disorder.

Public Significance Statement.

This article proposes that aspects of executive functioning may play a role in the onset and maintenance of avoidant/restrictive food intake disorder (ARFID), although this notion is largely untested by existing research. Further research on the role of executive functioning in ARFID may assist with refining models and treatments for this heterogeneous disorder.