Cognitive inflexibility has received considerable attention as a putative mechanism that might increase the risk for eating disorder (ED) onset, and influence illness course and treatment response. However, research findings have been mixed and have not led to advances in treatment outcome. Using a framework informed by cognitive neuroscience, we contend that separating two discrete facets of cognitive flexibility that are conflated in measures currently employed in the EDs field (i.e., attentional set-shifting and reversal learning) will help elucidate the mechanisms that underlie its relation with ED symptoms. Moreover, we argue that research that incorporates measures of attentional set-shifting and reversal learning will promote understanding of the role of cognitive flexibility in the expression, course, and treatment of EDs.
Defining Cognitive Flexibility: Attentional Set-Shifting and Reversal Learning
A critical step in studying cognitive flexibility is to clarify the behavioral and neural dimensions that are most salient to EDs. This is challenging because research in EDs has relied primarily on multidimensional clinical neuropsychological measures such as the Wisconsin Card Sorting Test (WCST) and Brixton Test, and there is considerable diversity in the assessments that have been employed. For example, many studies have used perseverative errors on the WCST as an index of cognitive inflexibility in EDs. However, because the WCST is a complex task that requires respondents to match cards under changing rules and feedback about performance, perseverative errors indicating cognitive inflexibility can reflect disruptions in two discrete neurocognitive processes: (1) attentional set-shifting, i.e., the ability to shift attention away from one abstract stimulus dimension (e.g., color) toward another (e.g., shape); and (2) reversal learning, i.e., the ability to override a recently acquired stimulus-reinforcement association (e.g., matching based on color) to apply a new stimulus-reinforcement association (e.g., matching based on shape).
Importantly, attentional set-shifting and reversal learning have distinct neural correlates (Table 1). For example, in animals, inactivations or lesions of the orbitofrontal cortex (OFC), but not the lateral prefrontal cortex, influence reversal learning, whereas inactivations or lesions of the lateral (primate) or medial (rodent) prefrontal cortex, but not the OFC, affect attentional set-shifting. In humans, reversal learning has been linked to function in the OFC and ventral striatum (VS), while attentional set-shifting has been related to the ventrolateral prefrontal cortex (VLPFC). Finally, there are distinctions in the neurochemical correlates of these dimensions, with depletion of prefrontal cortical dopamine (DA) and norepinephrine disrupting attentional set-shifting, and striatal DA, in particular at DA D2 receptors, and serotonin in the OFC influencing reversal learning. Thus, separating the behavioral components of attentional set-shifting and reversal learning, and their neural correlates, is crucial for advancing research on cognitive flexibility in EDs.
Table 1. Selected neural correlates of attentional set-shifting and reversal learning
Mediating Neural Structures
•Ventrolateral prefrontal cortex (VLPFC)
•Norepinephrine (NE) in the prefrontal cortex (PFC)
•Anterior cingulate cortex (ACC)
•Dopamine (DA) in the PFC
•Posterior parietal & temporal areas
•DA D1 receptors
•Orbitofrontal cortex (OFC)
•Serotonin (5-HT) in the OFC
•Ventral striatum (VS)
•DA in the striatum
•DA D2 receptors
How Should Attentional Set-Shifting and Reversal Learning be Discriminated and Measured?
The selection of behavioral tasks is critical for explicating the roles of attentional set-shifting and reversal learning in EDs. Unlike clinical neuropsychological measures that tap multiple neurocognitive domains, experimental behavioral tasks are designed to assess specific cognitive processes. For example, the intradimensional/extradimensional shift task from the Cambridge Neuropsychological Test Automated Battery was developed as an experimental analogue of the WCST to separate attentional set-shifting and reversal learning. During this nine-stage task, respondents are presented with pairs of stimuli and must learn to select the correct stimulus based on feedback following each trial. The rule for correct responding is modified at the beginning of each stage to dissociate aspects of cognitive flexibility. For example, the intradimensional shift stage examines rule generalization when novel stimuli are introduced, whereas the extradimensional shift stage examines the ability to inhibit or shift attention away from previous relevant stimulus dimensions (i.e., attentional set-shifting). A functional imaging paradigm based on the intradimensional/extradimensional shift task has been developed, and has been shown to engage the VLPFC and related set-shifting circuitry. Extradimensional shift errors on this task have been linked to numerous psychiatric disorders. However, research using the intradimensional/extradimensional shift task in ED patients has failed to document impairments in extradimensional set-shifting, which might suggest that attentional set-shifting is less salient to the expression of cognitive inflexibility than reversal learning.
Studies that assess attentional set-shifting and reversal learning separately within the same sample will help to clarify the relative roles of these processes in EDs. In humans, reversal learning frequently is measured using probabilistic reversal learning paradigms in which respondents learn to select the correct stimulus based on feedback, but some of the feedback is inaccurate. For example, during the first stage of a two-stage probabilistic reversal learning task, respondents are rewarded for selecting stimulus 1 and punished for selecting stimulus 2 on 80% of the trials; however, false feedback is delivered on 20% of the trials such that respondents are punished for selecting stimulus 1 and rewarded for selecting stimulus 2. In the second stage of the task, the probabilities are reversed. Probablistic reversal tasks are preferred over non-probabilistic tasks (e.g., reversal errors on the intradimensional/extradimensional shift task) because they are more difficult and more likely to encourage perseverative behavior after contingency reversals. Probabilistic reversal learning paradigms also have been developed for use in functional imaging studies, and have been shown to elicit responses in the VS and OFC. However, no study to our knowledge has used a probabilistic reversal learning task in a sample of individuals with EDs.
Using Measures of Attentional Set-Shifting and Reversal Learning to Advance Research on Cognitive Flexibility in EDs
Future work is needed to clarify whether attentional set-shifting is salient to EDs, or whether other forms of cognitive inflexibility, such as impaired reversal learning, are responsible for clinical neuropsychological findings. Hypotheses warranting further exploration include the possibility that attentional set-shifting is relevant only to particular forms of ED, or that differences between ED patients and controls in the neural correlates of attentional set-shifting exist despite similar task performance. Below we suggest three lines of research that could be advanced through the use of behavioral and neural measures of attentional set-shifting and reversal learning.
Phenotypic Heterogeneity in EDs
Differences in the psychological and biological correlates of anorexia nervosa-restricting type, anorexia nervosa-binge-eating/purging type, and bulimia nervosa raise the intriguing hypothesis that these presentations might be differentially associated with attentional set-shifting and reversal learning (Fig. 1).[4, 5] As such, administering behavioral measures of attentional set-shifting and reversal learning to individuals across the ED spectrum, and conducting imaging studies using conceptually relevant neurocognitive probes, could help elucidate phenotypic heterogeneity in ED presentation. Variations in the salience of attentional set-shifting and reversal learning also might help to explicate other forms of systematic heterogeneity in EDs, such as undercontrolled, overcontrolled, and low psychopathology presentations.
Cognitive Flexibility in Adolescents with Eating Disorders
Cross-sectional studies using behavioral measures have failed to document cognitive inflexibility in adolescents with EDs, leading some scholars to speculate that cognitive inflexibility is a consequence of disordered eating. However, a compelling alternative explanation for these findings is that individuals early in the course of an ED may be compensating for disrupted processing during task performance by: (a) increasing the magnitude of response in neural circuits underlying attentional set-shifting and reversal learning in order to successfully complete cognitive flexibility tasks; (b) recruiting additional neural structures to offset inadequate activity in neural circuits underlying attentional set-shifting and reversal learning; or (c) both. In support of this idea, research in non-ED-related psychiatric disorders and older adults has found alterations in neural processing during cognitive tasks even when task performance has been equal in the patient and control groups. Thus, studying the neural correlates of attentional set-shifting and reversal learning in adolescents with EDs is crucial to explicating the role of cognitive inflexibility in the early course of these disorders. Using more precise behavioral tasks, as opposed to multidimensional clinical neuropsychological measures, also will help to clarify which aspects of cognitive inflexibility are most salient to adolescent EDs. Finally, examining changes in behavioral and neural facets of attentional set-shifting and reversal learning over time will help elucidate the role of cognitive flexibility in the early progression of ED symptoms.
Ultimately, a better understanding of the behavioral and neural facets of cognitive flexibility with relevance to EDs will promote development of novel interventions and prevention programs for high-risk groups. For example, several medications have been shown to improve attentional set-shifting in individuals with psychiatric disorders, and animal research has found that second-generation antipsychotic agents attenuate reversal learning deficits. Finally, one limitation of cognitive remediation therapy is that it targets a broad range of neurocognitive functions that may or may not be salient to the expression or maintenance of EDs. In contrast, cognitive retraining strategies that focus on specific processes that have relevance to particular psychiatric presentations (e.g., attention bias modification for anxiety disorders) have been shown to produce significant decreases in psychiatric symptoms when compared to control interventions. Cognitive retraining interventions also can be used to help individuals accommodate areas of dysfunction, even if underlying deficits do not change. Thus, clarifying the processes that underlie cognitive inflexibility in EDs will facilitate the development of novel cognitive retraining strategies designed to target specific mechanisms that have salience to particular ED symptoms.