Skip to main content

Understanding implicit and explicit learning in adolescents with and without anorexia nervosa

Abstract

Background

Cognitive disturbances such as impairments in learning are thought to play a role in adult Anorexia Nervosa (AN). It is remains unclear to what extent these disturbances result from starvation of the brain, or relate to an abnormal premorbid cognitive profile. This study investigates learning processes in adolescents with AN, hypothesizing that implicit learning is intact, as found previously in explicit learning tasks. Secondly, we hypothesized that anxiety and depression symptoms, inherent to AN, are associated to learning processes in AN.

Methods

In total 46 adolescents diagnosed with AN and 44 control participants were administered an implicit category learning task in which they were asked to categorize simple perceptual stimuli (Gabor patches) based on a linear integration (i.e., an implicit task) of orientation and spatial frequency of the stimulus. A subgroup of adolescents (n = 38) also completed a task assessing explicit learning.

Results

Model-based analyses indicated that adolescents with AN performed significantly more accurately compared to their healthy peers regardless of whether they used the optimal strategy or not. Depression and anxiety did not relate to learning performance in the AN group.

Conclusions

Overall, our findings of augmented implicit and explicit learning in adolescents with AN corroborate recent studies that suggested higher stimulus-response learning during prediction error paradigms. Learning disturbances in adult AN may then be at least partly due to long-term malnourishment, highlighting the importance of early recognition and refeeding in treatments for AN.

Plain English summary

We know that some adults with anorexia nervosa (AN) experience difficulties in learning processes. It is remains unclear to what extent these difficulties result from long-term starvation of the brain. This study looked at learning processes in adolescents with AN who have a relatively short duration of illness. We also investigated whether anxiety and depression affected learning. Forty-six adolescents diagnosed with AN and 44 control participants completed tasks that assessed learning and questionnaires assessing depression and anxiety. We found that adolescents with AN performed more accurately compared to their healthy peers. However, depression and anxiety did not relate to learning performance. Overall, our findings suggest that individuals with AN, relative to their healthy peers, may be quicker in forming automatic responses and behaviors to cues. Learning disturbances in adult AN may then be at least partly due to long-term malnourishment, which highlights the importance of early recognition of AN and refeeding in treatments for AN.

Background

Anorexia nervosa (AN) is a severe psychiatric disorder with the highest mortality rates across mental disorders [1]. AN is characterized by extreme restriction of intake or purging of food, fear of gaining weight and disturbed experience of body weight or shape (DSM-5, [2]). The lifetime prevalence of AN among women is up to 4% [3] with a crude mortality rate of approximately 5% per decade [1, 2]. This debilitating disorder most typically develops during adolescence or young adulthood [4] and research suggests that prepubertal and early adolescent onset of AN may be on the rise [5].

Whilst available treatments for adolescents, in particular family-based interventions, are associated with good rates of remission (approximately 70% [6];) there is still a subgroup of young people who do not benefit from this treatment. lLittle is known about factors contributing to a more chronic prognosis [6, 7].

Recently it has been suggested that the focus for AN treatment should shift from mainly treating physical symptoms (i.e weight loss), and psychiatric symptoms (i.e., depression), to potentially underlying pathologies, such as disturbed cognitive processes, which have been described in adults with AN [8]. To further advance this direction, we investigated specific learning processes and whether these are comparable in younger AN patients to their peers without AN. Findings will contribute to unravelling whether impairments in cognitive processes such as learning are implicated in the development of AN, or whether these impairments are related to chronic starvation of AN. Subsequently, this knowledge will inform treatment directions.

Over the last few years studies have shown that some individuals with AN experience impairments in cognitive functioning. Studies in adults with AN show problems across a wide range of cognitive domains, such as, motor inhibition, visual processing speed, central coherence, visual–spatial ability, attention, learning and memory as well as decision making and cognitive flexibility [9, 10]. Currently it remains unknown however to what extent these difficulties are contributing to the development and maintenance of AN, or in turn, to what extent the chronic underweight of AN causes these cognitive impairments. Although longitudinal studies are the desirable method for answering these questions, these types of studies are costly and hindered by high levels of attrition in AN [11]. An alternative approach to gaining insight into the relation between cognitive functioning and AN is to study patients with a relative short duration of illness (i.e., adolescents) to compare with results in older samples. Seeing that the common age of onset of AN is early to mid-adolescents, studying young people with AN may provide important information about cognitive disturbances in AN at younger age [5].

Interestingly, published data on cognitive functioning in adolescent patients with AN posit a more mixed picture compared to the adult literature. Many studies, commonly including neuropsychological instruments, report no deficits at all or only subtle impairments in e.g. nonverbal intelligence functions and cognitive flexibility impairments such as audiomotor responses, and set-shifting abilities (e.g. [12,13,14,15,16,17,18]). In terms of general cognitive functioning, Schilder et al. [19] found that IQ was in fact higher in adolescent AN patients then the norm which suggests a superior cognitive functioning compared to peers.

Looking specifically at learning in AN there is less literature available. One increasingly popular hypothesis, based on recent advances in cognitive neuroscience, posits that persistent AN behaviors may be understood as maladaptive habits, which are driven by abnormal learning processes [20, 21]. This neurobiological “habit model of AN” [22, 23] suggests that for AN patients, eating behaviors become automatic responses very quickly and that little effort is needed to maintain these behaviors. On the other hand, discontinuing these dysfunctional habits becomes very difficult, as expressed in the often unsuccessful treatment of AN. In other words, stimulus-response learning may be augmented in individuals with AN.

Another important type of learning is category learning, which refers to the ability to make adaptive responses across a wide variety of situations and as such is a fundamental decision making process. Two separate but overlapping learning systems that contribute to category learning are the explicit and implicit learning systems. Explicit learning involves conscious learning, including (sets of) rules and feedback processes (rule-based learning) [24].

On the other hand, implicit learning refers to extracting predictive relationships in the form of statistical regularities or sequence of events from the environment without putting conscious effort into the process or even realizing the learning process at all (procedural-based learning [24]. The two types of learning are related to different brain areas and neural pathways, whereby explicit learning involves the hippocampal and medial temporal areas, whilst implicit learning engages frontal cortico-striatal circuits [25].

Explicit category learning involves both initial acquisition learning and updating explicitly-learned associations. This latter learning aspect is partly determined by a cognitive process called set-shifting, i.e. being able to shift attention between one task and another, whereby poor set-shifting interferes with being able to successfully update these explicitly-learned associations. In recent years set-shifting has gained a lot of attention in AN. While the literature suggests impaired set shifting in adults with AN [26], findings related to set-shifting in adolescent AN samples are mixed and whilst some studies show set-shifting impairments, other studies find that adolescents with AN perform on equal measure to HC groups (for a review see [15]). Mixed findings may be attributable to the complexity of different paradigms assessing different components of set-shifting. One task that has been previously used in studies investigating set-shifting and, more broadly, the ability to update explicitly-learned associations is the Houses and Castle task [27]. Set shifting deficits have been found in individuals with AN and individuals weight restored from AN [18, 28]. Individuals weight restored from AN also displayed hyper-learning, defined as a steeper learning curve, and learned the rules of the task faster than their healthy counterparts. This learning slope, however, was not significantly associated with the shift cost [28]. As far as we know, there are no studies explicitly examining hyper-learning in adolescents with AN. Following the theory that an amplification in learning processes, for example habit learning, is characteristic of AN, arguably then we may expect to find altered learning processes in adolescents with AN. Research on implicit category learning on the other hand is scarce. In fact, to our knowledge only one study looked at implicit learning in adolescents with AN [29]. Firk and colleagues [29] studied an adolescent sample before and after weight gain and found implicit sequence learning, which refers to learning the order of a sequence of stimuli, which is thought to be random, to be impaired, and that this was related to lower BMI. Looking at the adult literature, Shott et al. [30] found that in adults with AN, implicit category learning, which refers to learning how to categorize stimuli according to an unknown and non-verbalizable rule, was impaired. Other studies in adults showed (implicit) attention interferences for food-related words in individuals with patients with AN, but no implicit memory bias [31].

Furthermore, Shott et al. [30] found that implicit category learning was related to heightened novelty-seeking and lower sensitivity to punishment in adults with AN, which hints at the potential association with reward processes. Studies in adults show that the reward-related dopamine system is indeed implicated in cognitive functioning (e.g., reinforcement learning) [32]. Moreover, alterations in dopamine system activity has been associated with depression [33] and anxiety traits [34], both of which are pertinent to AN [34,35,36]. All of the studies above include adult samples; studies including adolescent samples are scarce. There is some evidence from cerebrospinal fluid and neuroimaging studies that the dopaminergic (DA) system is abnormal in adults ánd adolescents with AN, studies are lacking that directly linked DA function to behavior in AN [37,38,39]. The DA system is involved in Pavlovian model free learning, as well as habit and goal directed learning [40]. Elevated brain response during reward prediction error tasks may indicate altered Pavlovian stimulus-response learning in AN [38]. However, the interactions between the DA system and learning in AN needs further study. Nevertheless, it has been speculated that plasticity of brain DA function in adolescents is higher than in adults, and that this more flexible DA response may protect from DA-related learning inefficiencies [18]. It is therefore possible that whilst adults with AN display impaired learning, adolescents with AN will have intact learning due to age-dependent greater flexibility of their learning circuitry. Alongside the possible influence of the reward system, other potentially important effects on attention-dependents tasks include age-related, motivational and mood-related effects. For example, it has been suggested that those with a relatively short duration of AN (either adult or adolescent) may have ways to compensate poor cognitive functioning by activating more neural activity, or different neural circuits, or both [41]. High levels of perfectionism and fear for making mistakes, often reported by both adults and adolescents with AN [42] may also serve as a compensatory feature contributing to a performance comparable to healthy peers despite perhaps poorer cognitive functioning. With regards to mood disturbances there is evidence suggesting that both depressive and anxious symptomatology contribute to poorer performance on cognitive tasks in adults with AN [43, 44]. Whether depressive and anxious symptomatology contribute to implicit learning in adolescents with AN is another unexplored area. Recent studies have, for instance, indicated that individuals who score high on intolerance of uncertainty perform poorly on threat extinction (as assesses with responses to uncertain auditory stimuli that varied in threat level), which may contribute to suboptimal learning [45]. Intolerance of uncertainty is an important anxiety-related factor and refers to approach and avoidance responses to uncertainty whereby those with higher intolerance levels, are more likely to interpret ambiguous stimuli as more threatening, and have less confidence in their decisions in ambiguous situations [46]. Intolerance of uncertainty has been associated with the reward system [47] and may be an important factor for understanding learning processes in AN seeing that it is pertinent to both adults with AN [48,49,50] and adolescents [51, 52].

Depression has also been associated with altered learning and specifically in reward related context [53]. Our understanding of what brain regions and neurotransmitter systems are involved is still limited and again studies including adolescent samples are lacking However, several factors could play a role. Anxiety and depression as well as AN are associated with elevated cortisol levels as a sign of high stress, which could interfere with cognitive flexibility and learning [54]. Stress has been found to alter dopamine and noradrenaline circuitry and thereby altering working memory function and learning [55]. In sum, whilst there is some evidence for altered learning processes in adults with AN, and associations with affective states, studies investigating learning processes in adolescents with AN are scarce. Building on the theory that altered learning processes are characteristic of AN, we expect to find these alterations in adolescents as well.

Aims & hypotheses

The aims of this study were twofold. We wanted to test the hypothesis that implicit learning is intact in adolescent AN, similarly to explicit learning studies, as this may provide insight into the development of cognitive functioning from childhood years to adulthood and may shed some light onto the relation between learning abnormalities and the (long-term) neurobiological starvation effects in AN. Second, we wanted to test whether depressive and anxiety symptoms are related to worse learning performance in adolescents with AN.

Method

Participants

A total of 90 adolescent participants (11–17 years old) were recruited from two different sites (the Netherlands - NL, United States of America - USA), which will be described here separately. No participants were excluded in the USA groups; 3 participants were excluded in the NL groups.

NL sample

Twenty adolescents with a current diagnosis of AN according to DSM 5 criteria were recruited from a Dutch specialized Eating Disorders center (AN-NL group). Diagnoses were established by psychiatrists or clinical psychologists and confirmed with the Eating Disorder Examination (EDE [56];). Participants were excluded in the case of alcohol and drug abuse, history of or current diagnoses of other psychiatric disorders such as dementia, schizophrenia or mental retardation, and current diagnoses of diabetes, or a neurological disorder. Of these 20, 6 were taking anti-psychotics and 1 was taking mood-stabilizers. None were taking anti-depressants or sedatives. Eighteen healthy control adolescents were recruited in the Utrecht (NL) area through local advertisement flyers posted in a number of high schools, sports clubs and community centers (HC-NL group). They were included if they had no history of neurological medical diagnoses that may affect cognitive functioning, and no first-line relatives with a diagnosis of an eating disorder. Before participation, the experimenter completed the Mini International Neuropsychiatric Interview (M.I.N.I.: [57]; Dutch version: [58]), in order to screen for any possible (undiagnosed) psychiatric disorders. If there was any indication of an (undiagnosed) disorder (as seen from any of the subsections of the M.I.N.I.) participants were excluded from the study (n = 3).

USA sample

Twenty-six adolescents with a diagnosis of AN (AN-USA group) were recruited through an Eating Disorders program at a children’s hospital and a specialized Eating Disorder center (USA). All participants met DSM 5 criteria for AN (n = 24) or broadly defined AN (restricting atypical AN, n = 2) at the time of enrolment of the study. All individuals with AN completed the Clinical Diagnostic Interview Schedule for Children 4.0, to assess all major psychiatric diagnoses [59]. Participants were excluded if there was any indication of current substance use or other psychiatric disorders including dementia, mental retardation, schizophrenia or any neurological disorder. Those with diagnoses of anxiety and depression were included. Of these 26, 10 were taking anti-depressants and 4 were taking anti-psychotics. None were taking mood-stabilizers or sedatives. Twenty-six adolescent non-AN controls (HC-USA group) were recruited through local advertisements in the Denver metropolitan areas (USA). They completed the Clinical Diagnostic Interview Schedule for Children 4.0, to rule out any current or previous psychiatric disorders [59]. Non-AN controls had a lifetime history of body weight between 90 and 110% of ideal body weight since menarche.

Clinical measures

The NL-AN group’s BMI was assessed before participation, by measuring weight on a digital Tanita scale (Tanita Cooperation of America, Inc., Arlington Heights, IL) and height with a stadiometer. The NL-HC group’s BMI was assessed by asking participants to state their height and weight. All BMI were then calculated as kg/m2.

The USA-AN group’s BMI (kg/m2) was obtained from their hospital chart (weight was measured on a digital scale daily). The weight date was on the day of the testing session, which was between 1 and 2 weeks into treatment. The USA-HC group’s BMI (kg/m2) was assessed immediately before the testing session by weighing them on a digital Detecto scale (Detecto, Webb City, Missouri) and measuring their height with a Seca stadiometer.

All four groups of participants (NL-AN, NL-HC, USA-AN and USA-HC) were asked to fill out two questionnaires; the Children’s Depression Inventory to measure possible depressive symptoms (CDI: [60], α = 0.92; Dutch version: [61], α = 0.88 and a scale measuring intolerance of uncertainty (IUS: [62], α = 0.96; Dutch version: [63], α = 0.93). We used the Intolerance of Uncertainty Scale (IUS) to assess symptoms of anxious pathology. Intolerance of uncertainty is a key component of anxiety, and a wealth of evidence shows the contribution of intolerance of uncertainty to anxiety across a wide range of anxiety disorders and other psychological disorders (see a recent review by [64]).

Implicit category learning task

All 908 participants were asked to do an implicit category learning task, as previously used in Shott et al. [30]. In this task, participants were presented with Gabor patches (see Fig. 1 for examples), which they were asked to categorize into one of two categories (A and B). Each Gabor patch was presented until the participant’s response was made (“z” and “/” keys for categories A and B, respectively). After this, the participant received feedback for 1 s: the screen displayed the words “correct” or “wrong” respectively. Immediately after 1 s of feedback, a mask was displayed for 5 s in order to prevent participants from responding to the after-image of the previous stimulus. Then, the next trial began. The rule, unknown to participants, by which the Gabor patches had to be categorized, required a linear integration of two stimulus dimensions (spatial frequency and orientation of the lines in the stimulus). In each testing session, each of the presented stimuli was unique in its combined spatial frequency and orientation dimensions. For each testing session there were 80 trials, for which an equal number of Gabor patches from category A and B were generated randomly by sampling from two bivariate normal distributions (as originally done by [65]). Each Gabor patch was generated using MATLAB routines from Brainard’s [66] Psychophysics Toolbox.

Fig. 1
figure1

Examples of Gabor patch stimuli, image taken from Shott et al. [30]

The NL groups performed this task on a computer with a 15.4″ screen with a 1680 × 1050 resolution. The Gabor patches were thereby approximately 5 cm in height and at an approximate viewing distance of 45 cm, they subtended a visual angle of about 6.4°. The USA groups performed this task on a computer with a 21″ screen with a 1360 × 1024 resolution. Each Gabor patch was thereby 7 cm in diameter, which subtended a visual angle of about 8.9° from an approximate viewing distance of 45 cm.

This paradigm has been used extensively to gain a better understanding of the underlying processes in category learning in both normal and patient populations [67, 68].

Explicit category learning task

A selection of our sample (AN-NL and HC-NL; total n = 38) additionally completed an explicit category learning task, in addition to the implicit category learning task, i.e. the Houses & Castles task [18]. In this task, participants were randomly categorized into two groups: Houses group and Castles group. In each trial, participants were presented with either a cartoon image of a house or a castle (see Fig. 2 for examples), depending on their group, which was presented until the participant made a response. Each stimulus belonged to a category (“A” or “B”) based on an unknown rule. Participants were asked to categorize the stimuli by pressing a key (“z” key and “/” key for categories “A” and “B” respectively). Immediately after the participant’s response, feedback was shown for 0.75 s: displaying either the word “correct” or “wrong” beneath the image of the stimulus. This was followed by 1 s of blank screen, after which the next trial began. There was a total of 160 trials. Four dimensions with binary values could differ per stimulus per trial: castle stimuli – shape of foundation (diamond or square), location of ramparts (above or sunken into walls), number of rings around castle (one or two), color of drawbridge (yellow or green); house stimuli – color of door (red or blue), lighting inside window (light on or off), shape of roof (flat or triangular), type of plant (shrub or tree). During the first 80 trials, the rules for categorization were as follows: castle stimuli – shape of foundation (category “A”: square, category “B”: diamond); house stimuli – shape of roof (category “A”: flat, category “B”: triangular). During the last 80 trials, the rules for categorization were as follows: castle stimuli – number of rings (category “A”: one, category “B”: two), house stimuli – type of plant (category “A”: tree, category “B”: shrub). Participants were never informed of the rule shift and had to infer all rules from the provided feedback. Participants were given feedback on every trial and the contingencies were the same in each trial.

Fig. 2
figure2

Examples of castle and house stimuli, image taken from Shott et al. [18]

This task has been used many times before to reliably test set shifting/explicit and implicit category learning across normal and patient populations [18, 27].

Participants performed this task on a computer with a 15.4″ screen with a 1680 × 1050 resolution. Stimuli were approximately 6.5 cm in height and at an approximate viewing distance of 45 cm, they subtended a visual angle of about 8.3°. The stimuli were generated using MATLAB routines from Brainard’s [66] Psychophysics Toolbox.

Procedure

This study was approved by both appropriate USA and Dutch (medical) ethical committees. All participants and their parents or legal guardians gave consent for participation in this study. USA data was collected between February 2012–February 2014. An additional Dutch sample was included between October 2014 and August 2016 in order to 1) increase power for the implicit learning task; and 2) build on the [18] study by analyzing associations between the explicit learning processes and intolerance of uncertainty.

Primary analyses included a between-subjects design, where all participants were asked to do an implicit category learning task, and a sub-group of participants (NL-AN and NL-HC groups only) were asked to additionally do an explicit learning task.

In the case of participants who were not administered the additional explicit learning task (USA-AN and USA-HC groups), they were asked to complete the implicit category learning task at the beginning of the testing session. In the case of the other sub-group of participants (NL-AN and NL-HC groups), they were asked to first perform the explicit learning task, followed by the implicit learning task after a small break. This order was chosen as the explicit learning task is the easier one of the two so we expected that participants would thereby stay motivated enough after the first task to complete the second task.

All participants were asked to fill out all questionnaires at the end of the testing session. The experimenter stayed with the participant at all times during the testing session in case of fatigue, questions about the tasks or questionnaires, or in case of early termination of the experiment. Participants were not given compensation for their participation.

Statistical analyses

Statistical Package for the Social Sciences version 26 was used for the analyses. In order to see whether there were any significant differences in age or BMI between the AN adolescents and the non-AN controls, and between the USA and NL groups, independent samples t-tests were run with Bonferroni corrections for multiple testing. Sphericity as well as homogeneity of variances were checked and corrected for accordingly at all times. Where sphericity could not be assumed within an ANOVA, the Greenhouse-Geisser results are reported. Estimates of effect size are calculated using partial eta squares or Cohen’s D [69], where .2 = small effect, .5 = medium effect and .8 = large effect.

Learning outcomes

Model-based analyses implicit learning task

Using mathematical models, this task allows for insight into the specific approach participant use when learning the task [67, 70, 71]. As explained by Shott et al. [30], these models can identify AN patients who adopted a procedural-based approach to learning compared to healthy controls, in order to assess impairments in procedural-based learning in patients. Two classes of models will be compared, namely the procedural-based (PB) approach, and the hypothesis-testing (HT) approach. The optimal PB model assumes that participant used the rule displayed in Fig. 1 as the solid line. The second PB model was the general linear classifier (GLC), which also assumes that the participant’s decision on each trial is based on a linear integration. HT models assume that the participant set a criterion and that there were four response regions: low frequency/ shallow angle, low frequency/steep angle, high frequency/shallow angle and high frequency/ steep angle (for a detailed explanation of these models see Shott et al. [30].

Statistical analyses

Following procedures as described by Shott et al. [30], to explore differences in implicit learning performance, a 2 (group) × 4 (block) mixed-design ANOVA (to compare AN to HC) was run with the following measures: 1) accuracy (number of correct responses divided by number of trials), 2) reaction time (RT, in seconds) and 3) reaction time variability (standard deviation of reaction time). Moreover, 4) a learning curve (accuracy in block 4 – accuracy in block 1) was computed and 4-way ANOVA’S were used to examine group differences. Post-hoc tests were examined to ensure no country group differences (NL versus USA AN groups).

Following procedures from Shott et al. [30], for the explicit learning task, to explore differences in accuracy (number of correct responses divided by the number of trials) in the explicit learning task, a 2 (group) × 8 (block) mixed-design ANOVA was run (in the Dutch samples only).

Learning outcomes and anxious and depressive symptomatology

To explore associations between implicit learning outcomes and anxious and depressive symptoms Pearson’s correlation analyses were run in all four groups including, depression, BMI age, and learning curve (implicit learning) variables (for an elaborate explanation on the using the learning curve within the correlation analyses, rather than the other outcomes we refer to procedures described by Shott & colleagues [18].

To explore associations between implicit learning outcomes, explicit learning, anxious and depressive symptoms, Pearson’s correlation analyses were run in AN (NL-AN only) and in HC (NL-HC only) separately including learning curve (implicit learning outcome), accuracy block 5 (explicit learning), intolerance of uncertainty, depression, BMI and age.

Results

Age and BMI

There were no significant difference in age between AN patients (NL versus USA (NL: mean = 15.60, SD =1.23; USA: mean = 14.73, SD = 1.56)) and non-AN controls (NL versus USA (NL: mean = 15.22, SD = 1.47; USA: mean 14.19, SD: 1.86)). There was a significant difference in age between the two USA and the NL groups whereby the USA groups were slightly younger than the AN-NL group (t(86.34) = − 3.73, p < 0.01, Cohen’s d = 0.79), but including age as a covariate did not change the results.

As expected a significant difference in BMI between ANs and HCs was found (see Table 1). Low BMI is inherent to AN diagnosis and was therefore not added into the analyses as covariate. No significant differences in BMI between the two AN groups or between the two HC groups were found.

Table 1 Age, BMI, depression, intolerance of uncertainty, implicit leaning (means, standard deviations and ranges; N = 90)

Implicit learning task

Due to the mixed design ANOVA on implicit learning data showing heterogeneity of variance on all measures (i.e. non-normal distribution), all data was logarithmically transformed to normalize the data and reduce heterogeneity of variances (according to [72]). A skewness analysis of the untransformed data showed that implicit learning data was indeed skewed (max skewness value = 4.78, SES = 0.25). The skewness observed in the logarithmically transformed data was improved as compared to the untransformed data, with all skewness values lying between − 0.54 and 0.36 (SES = 0.25), which is within the acceptable range of skewness (e.g. [73]). We therefore deemed the logarithmic transformation adequate to normalize the data. Another 2 (group) × 4 (block) mixed-design ANOVA was then run on the log10 transformations of accuracy, reaction time and reaction time variability. For a summary of all implicit learning task results, see Table 1.

Implicit learning task: accuracy

The ANOVA revealed a main effect of group, F(1,88) = 7.77, p = 0.01, ηp2 = 0.08, where the AN groups were overall more accurate than the HC groups (small effect). A main effect of block was found, F(2.30,201.92Footnote 1) = 20.59, p < 0.01, ηp2 = 0.19, where all groups improved across blocks. No significant interaction of block x group was found. For an illustration of the accuracy results, see Fig. 3. The NL and USA AN groups did not differ significantly from each other.

Fig. 3
figure3

Changes in accuracy (number of correct responses divided by the number of trials, log transformed) in implicit learning task across blocks, differentially between AN HC groups. Error bars show standard error

Implicit learning task: reaction time

No significant main effect of group on reaction time was found. There was a significant main effect of block on reaction time, F(2.58,227.17) = 28.54, p < 0.01, ηp2 = 0.25, where participants’ reaction times decreased across blocks. No significant group x block interaction was found. The NL and USA AN groups did not differ significantly from each other.

Implicit learning task: reaction time variability

For the reaction time variability, a significant main effect of group was found, F(1,88) = 5.51, p = 0.02, ηp2 = 0.06, where the AN group showed less variability in reaction times than the HC group. A significant main effect of block was found, F(2.31,203.22) = 25.32, p < 0.01, ηp2 = 0.22, where all participants’ variability decreased over time. No significant block x group interaction was found. The reaction time variability results are displayed in Fig. 4. The NL and USA AN groups did not differ significantly from each other.

Fig. 4
figure4

Changes in reaction time variability (standard deviation of reaction time in seconds, log transformed) in implicit learning task across blocks, differentially for AN and HC group

Implicit learning task: learning curve

No significant differences in learning curve (accuracy in last block minus accuracy in first block) between the AN and HC groups were found. The NL and USA AN groups also did not differ significantly from each other.

Implicit learning task: model results

In line with Shott et al. [30], to determine whether the model-based subgroups differed, accuracy rates in the final block for the AN and HC participants who used either procedural-based (PB) or hypothesis-testing (HT) approach were contrasted (see Fig. 5). T-tests showed that for both the AN and HC participants accuracy for the PB approach was significantly better compared to the HT approach (AN: p < .01; HC: p < .05). Moreover, for both the PB and HT approach, the AN participants performed more accurately than the HC participants (p < .01, when controlling for depression, anxiety or medication p < .05).

Fig. 5
figure5

Accuracy results for the Hypothesis Tested (HT) versus Procedural Based (PB) method of learning (*p < .05)

Explicit learning task

No heterogeneity of variance was found for the planned ANOVA, therefore no transformation had to be performed on the set-shifting data. For a summary of explicit learning task results, see Table 2.

Table 2 Means, standard deviations and ranges of explicit learning task results in the NL groups (N = 38)

Explicit learning: accuracy

The ANOVA revealed a significant main effect of group, F(1,36) = 11.35, p = 0.01, ηp2 = 0.24, where the AN group was consistently more accurate than the HC group (large effect). A significant main effect of block was found, F(3.39,122.02) = 10.94, p < 0.01, ηp2 = 0.23. No significant block x group interaction was revealed. For a visualization of the explicit learning task accuracy results, see Fig. 6.

Fig. 6
figure6

Changes in accuracy (number of correct responses divided by number of trials) in the explicit learning task across blocks, differentially for both groups. Error bars show standard error

Explicit learning: shift costs

To explore whether there were any differences in shift cost (accuracy in block 5 minus accuracy in block 4) between the two groups, an independent samples t-test was run. To determine the impact of the actual shift, a Shift-Cost score was computed by subtracting each participant’s accuracy on block 5 from their accuracy on block 4 (higher scores equaled a greater shift-cost). Group differences indicate how well a particular group did, compared to the other, at coping with the rule change. No significant differences were found. To investigate this further, an independent samples t-test was run on accuracy in block 5, as well as accuracy in block 4, between groups. After Bonferroni corrections, it was found that accuracy in block 5 differed between groups, t(36) = 2.76, p = 0.02, Cohen’s d = 0.90, but not accuracy in block 4 (this only yielded a significant result before Bonferroni corrections, t(36) = 2.13, p = 0.04, Cohen’s d = 0.69). This suggests that participants were nearly as good as each other at the end of block 4, but dealt with the set change after the rule shift differently, yielding different accuracies in block 5.

Relationships between implicit learning, intolerance of uncertainty, depression, BMI and age

As there were significant differences between the USA and NL groups regarding the Intolerance of Uncertainty Scale and Children’s Depression Inventory in both the HC and AN groups 4 different analyses were run including learning curve, intolerance of uncertainty, depression age and BMI. In the USA-AN, NL-AN and NL-HC groups no significant correlations were found between any of the implicit learning outcomes and the clinical variables. In the USA-HC group a smaller learning curve was associated to a higher BMI (r = −.47, p = 0.02) and to a higher age (r = −.43, p = 0.03).

Relationships between implicit learning, intolerance of uncertainty, depression, BMI, age and explicit learning (NL groups only).

In the NL-AN group, IUS scores were significantly associated to explicit learning outcomes (Accuracy: r = .49, p = 0.03; Costs: r = −.46, p = 0.05), whereby higher intolerance of uncertainty was associated to higher SS accuracy and lower SS costs (i.e. more/stronger intolerance of uncertainty was related to better learning).

Discussion

In this study we aimed to understand implicit and explicit learning in adolescents with AN, and to explore associations between learning outcomes and anxious and depressive symptomatology. Interestingly, in terms of implicit learning, accuracy performance of AN participants was superior to that of the HC, and this was true for both the procedural-based (PB) and the hypothesis-testing (HT) model types. As expected, performance on the other implicit learning outcomes, reaction time and variability in these times, were comparable between the AN and HC participants. Similarly, for explicit learning, AN participants had higher accuracy rates compared to HCs. In both the combined AN and HC groups there were no associations between the implicit learning outcomes and clinical variables such as age, BMI, intolerance of uncertainty and depression. In the USA-HC group poorer implicit learning was associated to lower BMI and lower age, which may be due to developmental processes.

The finding of superior accuracy outcomes in AN on both tasks applied in this study is particularly interesting, seeing that a recent systematic review by Olivo et al. [74] concludes that on most cognitive domains, adolescents with AN are comparable to their peers in term of behavioral performance. Our finding of better performance compared to non-AN controls on a specific type of learning task adds new important information regarding cognitive functioning in adolescents with AN. Looking at learning more broadly, findings fit with the theory that individuals with AN may have augmented stimulus-response learning. As posited by the “habit model of AN” ([20]; Walsh, 2013), for individuals with AN, behaviors become automatic responses quickly and are then also maintained without much effort. Discontinuing these habits however is much harder, which may explain why some AN symptoms remain difficult to treat.

Furthermore, an earlier study showed that patients with AN had higher IQs than the population norm, which may also contribute to better performance [19]. Another possible explanation for more optimal behavioral performance may lie with high levels of perfectionism that may partly drive this overperformance [75]. That is, adolescents who develop AN put in more effort to “get it right”, which would reflect the high perfectionism commonly present in individuals with AN [42]. A limitation of the study however is that it did not include IQ or perfectionism measures so this remains speculation. Future studies on learning in AN should apply perfectionism scales and test this hypothesis.

It is also possible that adolescents with AN, with a (usually) shorter duration of illness are in a state of cognitive and perfectionistic overdrive, driven by a brain pathophysiology that is in a state of overexcitability and associated with high intellectual capacity. Interestingly, such an “overexcitable cognitive ability” has been associated with hyper-reactivity of the central nervous system [76], which is associated to a risk for psychopathology [77]. An important neural system implicated in cognitive functioning and perhaps explanatory for our results of better learning performance in adolescent AN is the dopamine system. Striatal dopamine pathways are involved in major cognitive domains such as feedback sensitivity, which in turn affects learning processes. Indeed, previous literature highlights alterations in feedback sensitivity, especially punishment sensitivity, in adults and adolescents with AN [37, 78] which may affect learning strategies [32]. Furthermore, brain dopamine circuitry is a major contributor to model free and model based learning, namely Pavlovian prediction error learning, habit learning and goal directed instrumental learning [40]. This study was not designed to test dopamine circuit function and thus does not allow testing for these hypotheses. We are currently planning future studies that will include biological markers of the dopamine system when investigating learning in AN.

Our results are in line with previous literature showing that cognitive processes that may be disturbed in adults with AN are intact in adolescents with AN [12, 13, 15, 16, 18, 30]. It is therefore possible that cognitive deficits in adults are at least partly contributable to the illness itself, which may be explained by the neurobiological effects of long-term starvation [79]. The initial hyper-drive in adolescent AN during the continuing duration of illness (and associated long-term malnourishment) then transfers into a state of burn-out (and associated cognitive problems) in adults with long-term AN (as theorized by [74, 80]). Indeed, Shott et al. [30] found that adults with AN still performed poorly on the same implicit learning task used in this study, even if they applied the correct model or strategy. Whether these suspected illness-related changes are permanent is an important question for further research to examine, yet the relation between cognitive functioning and malnourishment in AN is complex, and study findings aiming to entangle this relationship are mixed. For example, as highlighted by a recent review [81], whilst some literature suggests that cognitive difficulties in AN are related to low weight, other studies find that cognitive deficits were not or only partly associated with malnourishment in AN (see also [82]). Also data on the reversibility of these potential malnourishment-related cognitive deficits is mixed; some studies find that AN-related cognitive difficulties are reversable after weight gain [83], whilst others find that even short-term malnourishment can lead to irreversible brain changes in adolescents with AN (for a review see [84]). The lack of associations between learning outcomes and clinical variables in the AN group was unexpected, in particular seeing that some recent studies demonstrated negative effects of anxious and depressive symptoms on cognitive functioning in AN (i.e. social problem solving; [85]) and central coherence [86]. Due to group differences in scores on the clinical instruments analyses were run in the separate smaller groups. This may have resulted in a power issue which in turn may have resulted in non-significant associations that in fact could be significant with a larger sample. On the other hand, a recent meta-analysis concluded that depression is not associated to set-shifting in adults [10]. In line with this finding, a recent study including adolescents showed that despite higher levels of depression in the AN group, set-shifting ability did not differ between AN and healthy controls [16]. Of note, results of the current study show that intolerance of uncertainty was significantly higher in the AN adolescents than in the non-AN controls, and that stronger intolerance of uncertainty was related to better explicit learning confirming that intolerance of uncertainty may be an important clinical factor in adolescents with AN [52]. How and to what extent intolerance of uncertainty fits into AN pathology requires further examination in future studies. Whilst no differences between the NL and USA groups on learning outcomes were found, interestingly, the NL and USA groups did differ in terms of anxious and depressive symptomology, with the NL-AN group reporting more severe depression and intolerance of uncertainty. Whether this is indeed a cultural difference in terms of severity, or can be explained by other cultural differences (i.e. interpretation of questions) we can’t conclude from this study.

Whilst these results are promising, it is important to keep in mind that the observed effects are small and that due to relatively small groups, interpretation should be done with caution and replication studies are warranted, in particular the findings for the explicit learning task seeing that only the NL groups completed these (18 HC versus 20 AN = total n = 38). However, for the associations between learning outcomes and clinical variables, some correlation coefficients were in fact quite high, suggesting that in larger groups significant correlations may be detected. Further studies should include larger samples and it is recommendable to record (clinical characteristics of) those who were invited but refused to take part in order to avoid any participation biases. Future studies may also want to ensure matching the HC group to the patient groups to optimize comparability.

Although we did include age in this study, we did not include illness duration, which may well be an important factor seeing the effects of more chronic and long-term AN on the brain, and cognitive processes. Lastly, due to a technical error we had to rely on self-report weight data for some participants, future studies should ensure objective weight information. Moreover, studies aiming to understand the complex relation between BMI and cognitive processes in adolescents may benefit from using the WHO values (% median BMI).

Conclusions

Taken together these findings shed light on learning processes in adolescents with AN, in that learning appears intact, or even enhanced, compared to their peers. This may be an indication that cognitive difficulties such as impaired implicit and explicit learning in adults with AN may result from (enduring) starvation or other illness related factors. Future research should aim to examine the effects of acute and long-term malnourishment and weight gain on learning processes.. The better performance in the adolescent AN group is in line with other research examining cognitive functions in adolescents, and the opposite compared to research in adults. It is therefore possible that the brain activation in AN when young is in a form of hyper-learning, i.e. showing steeper learning curves then their peers without AN, and that this hyper-learning is in turn potentially driven by anxiety and perfectionism to result in excellent task performance [76, 77]. However, this state is not sustainable, food restriction may also take its toll on the brain, and this may eventually lead to poor performance in adults.

Taking into account studies that do find learning impairments in adolescents with AN, further research should focus on unravelling different learning processes and their underlying neurocircuits. This may have direct clinical implications in that identifying the underlying pathophysiology of altered category learning in AN may allow us to identify interventions that maintain normal cognitive flexibility from young to adult age, which could improve outcome when treating adults with AN.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Notes

  1. 1.

    As Mauchly’s test of sphericity was violated Greenhouse-Geisser corrected values are reported

Abbreviations

AN:

Anorexia nervosa

APA:

American Psychology Association

BMI:

Body mass index

CDI:

Children’s Depression Inventory

DA system:

Dopaminergic system

DSM-IV:

Diagnostic statistical manual-IV

EDE:

Eating Disorder Examination

HC:

Healthy control

HT approach:

Hypothesis-testing approach

IQ:

Intelligence quotient

NL:

Netherlands

IUS:

Intolerance of Uncertainty Scale

PB approach:

Procedural-based approach

References

  1. 1.

    Fichter MM, Quadflieg N. Mortality in eating disorders-results of a large prospective clinical longitudinal study. Int J Eat Disord. 2016;49(4):391–401. https://doi.org/10.1002/eat.22501.

    Article  PubMed  Google Scholar 

  2. 2.

    American Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-5®). American Psychiatric Pub; 2013.

  3. 3.

    Smink FR, Van Hoeken D, Hoek HW. Epidemiology of eating disorders: incidence, prevalence and mortality rates. Curr Psychiatry Rep. 2012;14(4):406–14. https://doi.org/10.1007/s11920-012-0282-y.

    Article  PubMed  PubMed Central  Google Scholar 

  4. 4.

    Campbell K, Peebles R. Eating disorders in children and adolescents: state of the art review. Pediatrics. 2014;134(3):582–92.

  5. 5.

    Petkova H, Simic M, Nicholls D, Ford T, Prina AM, Stuart R, et al. Incidence of anorexia nervosa in young people in the UK and Ireland: a national surveillance study. BMJ Open. 2019;9(10):e027339. https://doi.org/10.1136/bmjopen-2018-027339.

    Article  PubMed  PubMed Central  Google Scholar 

  6. 6.

    Couturier J, Kimber M, Szatmari P. Efficacy of family-based treatment for adolescents with eating disorders: a systematic review and meta-analysis. Int J Eat Disord. 2013;46(1):3–11. https://doi.org/10.1002/eat.22042.

    Article  PubMed  Google Scholar 

  7. 7.

    Marucci S, Ragione LD, De Iaco G, Mococci T, Vicini M, Guastamacchia E, et al. Anorexia nervosa and comorbid psychopathology. Endocr Metab Immune Disord Drug Targets (Formerly Curr Drug Targets Immune Endocr Metabol Disord). 2018;18(4):316–24.

    Article  Google Scholar 

  8. 8.

    Zipfel S, Giel KE, Bulik CM, Hay P, Schmidt U. Anorexia nervosa: aetiology, assessment, and treatment. Lancet Psychiatry. 2015;2(12):1099–111.

    Article  Google Scholar 

  9. 9.

    Lena SM, Fiocco AJ, Leyenaar JK. The role of cognitive deficits in the development of eating disorders. Neuropsychol Rev. 2004;14(2):99–113. https://doi.org/10.1023/B:NERV.0000028081.40907.de.

    Article  PubMed  Google Scholar 

  10. 10.

    Smith KE, Mason TB, Johnson JS, Lavender JM, Wonderlich SA. A systematic review of reviews of neurocognitive functioning in eating disorders: the state-of-the-literature and future directions. Int J Eat Disord. 2018;51(8):798–821. https://doi.org/10.1002/eat.22929.

    Article  PubMed  PubMed Central  Google Scholar 

  11. 11.

    Abdelbaky G, Hay P, Touyz S. Factors associated with treatment attrition in anorexia nervosa: a systematic review. OA Evid Based Med. 2013;1:1–3.

    Article  Google Scholar 

  12. 12.

    Bühren K, Mainz V, Herpertz-Dahlmann B, Schäfer K, Kahraman-Lanzerath B, Lente C, et al. Cognitive flexibility in juvenile anorexia nervosa patients before and after weight recovery. J Neural Transm. 2012;119(9):1047–57. https://doi.org/10.1007/s00702-012-0821-z.

    Article  PubMed  Google Scholar 

  13. 13.

    Calderoni S, Muratori F, Leggero C, Narzisi A, Apicella F, Balottin U, et al. Neuropsychological functioning in children and adolescents with restrictive-type anorexia nervosa: an in-depth investigation with NEPSY–II. J Clin Exp Neuropsychol. 2013;5(2):167–79.

    Article  Google Scholar 

  14. 14.

    Kjaersdam Telléus G, Jepsen JR, Bentz M, Christiansen E, Jensen SO, Fagerlund B, et al. Cognitive profile of children and adolescents with anorexia nervosa. Eur Eat Disord Rev. 2015;23(1):34–42. https://doi.org/10.1002/erv.2337.

    Article  PubMed  Google Scholar 

  15. 15.

    Lang K, Stahl D, Espie J, Treasure J, Tchanturia K. Set shifting in children and adolescents with anorexia nervosa: an exploratory systematic review and meta-analysis. Int J Eat Disord. 2014;47(4):394–9. https://doi.org/10.1002/eat.22235.

    Article  PubMed  Google Scholar 

  16. 16.

    Rößner A, Juniak I, van Noort BM, Pfeiffer E, Lehmkuhl U, Kappel V. Cognitive flexibility in juvenile anorexia nervosain relation to comorbid symptoms of depression, obsessive compulsive symptoms and duration of illness. Z Kinder Jugendpsychiatr Psychother. 2016; Dec 12.

  17. 17.

    Sarrar L, Ehrlich S, Merle JV, Pfeiffer E, Lehmkuhl U, Schneider N. Cognitive flexibility and agouti-related protein in adolescent patients with anorexia nervosa. Psychoneuroendocrinology. 2011;36(9):1396–406. https://doi.org/10.1016/j.psyneuen.2011.03.014.

    Article  PubMed  Google Scholar 

  18. 18.

    Shott ME, Filoteo JV, Bhatnagar KA, Peak NJ, Hagman JO, Rockwell R, et al. Cognitive set-shifting in anorexia nervosa. Eur Eat Disord Rev. 2012b;20(5):343–9. https://doi.org/10.1002/erv.2172.

    Article  PubMed  PubMed Central  Google Scholar 

  19. 19.

    Schilder CM, van Elburg AA, Snellen WM, Sternheim LC, Hoek HW, Danner UN. Intellectual functioning of adolescent and adult patients with eating disorders. Int J Eat Disord. 2017;50(5):481–9.

    Article  Google Scholar 

  20. 20.

    Davis L, Walsh BT, Schebendach J, Glasofer DR, Steinglass JE. Habits are stronger with longer duration of illness and greater severity in anorexia nervosa. Int J Eat Disord. 2020;53(5):683–9. https://doi.org/10.1002/eat.23265.

    Article  Google Scholar 

  21. 21.

    Walsh BT. The enigmatic persistence of anorexia nervosa. Am J Psychiatry. 2013;170(5):477–84.

  22. 22.

    Haynos AF, Hall LM, Lavender JM, Peterson CB, Crow SJ, Klimes-Dougan B, et al. Resting state functional connectivity of networks associated with reward and habit in anorexia nervosa. Hum Brain Mapp. 2019;40(2):652–62. https://doi.org/10.1002/hbm.24402.

    Article  PubMed  Google Scholar 

  23. 23.

    Steinglass JE, Walsh BT. Neurobiological model of the persistence of anorexia nervosa. J Eat Disord. 2016;4(1):1–7.

    Article  Google Scholar 

  24. 24.

    Reber AS. Implicit learning and tacit knowledge: essay on the cognitive unconscious (Oxford Psychology Series). Oxford: Oxford University Press Inc; 1993.

    Google Scholar 

  25. 25.

    Yang J, Li P. Brain networks of explicit and implicit learning. PLoS One. 2012;7(8):e42993. https://doi.org/10.1371/journal.pone.0042993.

    Article  PubMed  PubMed Central  Google Scholar 

  26. 26.

    Fuglset TS. Set-shifting, central coherence and decision-making in individuals recovered from anorexia nervosa: a systematic review. J Eat Disord. 2019;7(1):1–4.

    Article  Google Scholar 

  27. 27.

    Filoteo JV, Maddox WT, Ing AD, Zizak V, Song DD. The impact of irrelevant dimensional variation on rule-based category learning in patients with Parkinson's disease. J Int Neuropsychol Soc. 2005;11(5):503.

  28. 28.

    Filoteo JV, Paul EJ, Ashby FG, Frank GK, Helie S, Rockwell R, et al. Simulating category learning and set shifting deficits in patients weight-restored from anorexia nervosa. Neuropsychol. 2014;28(5):741–51. https://doi.org/10.1037/neu0000055.

    Article  Google Scholar 

  29. 29.

    Firk C, Mainz V, Schulte-Ruether M, Fink G, Herpertz-Dahlmann B, Konrad K. Implicit sequence learning in juvenile anorexia nervosa: neural mechanisms and the impact of starvation. J Child Psychol Psychiatry. 2015;56(11):1168–76. https://doi.org/10.1111/jcpp.12384.

    Article  PubMed  Google Scholar 

  30. 30.

    Shott ME, Filoteo JV, Jappe LM, Pryor T, Maddox WT, Rollin MD, et al. Altered implicit category learning in anorexia nervosa. Neuropsychol. 2012a;26(2):191–201. https://doi.org/10.1037/a0026771.

    Article  Google Scholar 

  31. 31.

    Hermans D, Pieters G, Eelen P. Implicit and explicit memory for shape, body weight, and food-related words in patients with anorexia nervosa and nondieting controls. J Abnorm Psychol. 1998;107(2):193–202. https://doi.org/10.1037/0021-843X.107.2.193.

    Article  PubMed  Google Scholar 

  32. 32.

    Cavanagh JF, Frank MJ, Allen JJ. Social stress reactivity alters reward and punishment learning. Soc Cogn Affect Neurosci. 2011;6(3):311–20. https://doi.org/10.1093/scan/nsq041.

    Article  PubMed  Google Scholar 

  33. 33.

    Ayano GJ. Dopamine: receptors, functions, synthesis, pathways, locations and mental disorders: review of literatures. J Ment Disord Treat. 2016;2(120):2.

    Google Scholar 

  34. 34.

    Lawford BR, Young R, Noble EP, Kann B, Ritchie T. The D 2 dopamine receptor (DRD2) gene is associated with co-morbid depression, anxiety and social dysfunction in untreated veterans with post-traumatic stress disorder. Eur Psychiatry. 2006;21(3):180–5.

    Article  Google Scholar 

  35. 35.

    Allen KL, Byrne SM, Oddy WH, Crosby RD. DSM–IV–TR and DSM-5 eating disorders in adolescents: prevalence, stability, and psychosocial correlates in a population-based sample of male and female adolescents. J Abnorm Psychol 2013;122(3):720, 732, https://doi.org/10.1037/a0034004.

    Article  PubMed  Google Scholar 

  36. 36.

    Lloyd EC, Haase AM, Verplanken B. Anxiety and the development and maintenance of anorexia nervosa: protocol for a systematic review. Syst Rev. 2018;7(1):1–5.

    Article  Google Scholar 

  37. 37.

    DeGuzman M, Shott ME, Yang TT, Riederer J, Frank GK. Association of elevated reward prediction error response with weight gain in adolescent anorexia nervosa. Am J Psychiatry. 2017;174(6):557–65. https://doi.org/10.1176/appi.ajp.2016.16060671.

    Article  PubMed  PubMed Central  Google Scholar 

  38. 38.

    Frank GK, DeGuzman MC, Shott ME, Laudenslager ML, Rossi B, Pryor T. Association of brain reward learning response with harm avoidance, weight gain, and hypothalamic effective connectivity in adolescent anorexia nervosa. JAMA Psychiatry. 2018;75(10):1071–80.

    Article  Google Scholar 

  39. 39.

    Kaye WH, Ebert MH, Raleigh M, Lake CR. Abnormalities in CNS monoamine metabolism in anorexia nervosa. Arch Gen Psychiatry. 1984;41(4):350–5. https://doi.org/10.1001/archpsyc.1984.01790150040007.

    Article  PubMed  Google Scholar 

  40. 40.

    Daw ND, O’Doherty JP. Chapter 21: multiple systems for value learning. In: Glimcher PW, Fehrs E, editors. Neuroeconomics. 2nd ed. San Diago: Academic Press; 2014. p. 393–410.

    Chapter  Google Scholar 

  41. 41.

    Wildes JE, Forbes EE, Marcus MD. Advancing research on cognitive flexibility in eating disorders: the importance of distinguishing attentional set-shifting and reversal learning. Int J Eat Disord. 2014;47(3):227–30. https://doi.org/10.1002/eat.22243.

    Article  PubMed  Google Scholar 

  42. 42.

    Lloyd S, Yiend J, Schmidt U, Tchanturia K. Perfectionism in anorexia nervosa: novel performance based evidence. PLoS One. 2014;9(10):e111697. https://doi.org/10.1371/journal.pone.0111697.

    Article  PubMed  PubMed Central  Google Scholar 

  43. 43.

    Andrés-Perpiña S, Lozano-Serra E, Puig O, Lera-Miguel S, Lázaro L, Castro-Fornieles J. Clinical and biological correlates of adolescent anorexia nervosa with impaired cognitive profile. Euro Child Adol Psych. 2011;20(11-12):541–9. https://doi.org/10.1007/s00787-011-0216-y.

    Article  Google Scholar 

  44. 44.

    Wilsdon A, Wade TD. Executive functioning in anorexia nervosa: exploration of the role of obsessionality, depression and starvation. J Psychiatr Res. 2006;40(8):746–54. https://doi.org/10.1016/j.jpsychires.2005.10.006.

    Article  PubMed  Google Scholar 

  45. 45.

    Morriss J, Saldarini F, Van Reekum CM. The role of threat level and intolerance of uncertainty in extinction. Int J Psychophysiol. 2019;142:1–9. https://doi.org/10.1016/j.ijpsycho.2019.05.013.

    Article  PubMed  Google Scholar 

  46. 46.

    Birrell J, Meares K, Wilkinson A, Freeston M. Toward a definition of intolerance of uncertainty: a review of factor analytical studies of the intolerance of uncertainty scale. Clin Psychol Rev. 2011;31(7):1198–208. https://doi.org/10.1016/j.cpr.2011.07.009.

    Article  PubMed  Google Scholar 

  47. 47.

    Nelson BD, Shankman SA, Proudfit GH. Intolerance of uncertainty mediates reduced reward anticipation in major depressive disorder. J Affect Disord. 2014;158:108–13. https://doi.org/10.1016/j.jad.2014.02.014.

    Article  PubMed  PubMed Central  Google Scholar 

  48. 48.

    Frank GK, Roblek T, Shott ME, Jappe LM, Rollin MD, Hagman JO, et al. Heightened fear of uncertainty in anorexia and bulimia nervosa. Int J Eat Disord. 2012;45(2):227–32. https://doi.org/10.1002/eat.20929.

    Article  PubMed  Google Scholar 

  49. 49.

    Kesby A, Maguire S, Brownlow R, Grisham JR. Intolerance of uncertainty in eating disorders: an update on the field. Clin Psychol Rev. 2017;56:94–105. https://doi.org/10.1016/j.cpr.2017.07.002.

    Article  PubMed  Google Scholar 

  50. 50.

    Sternheim L, Konstantellou A, Startup H, Schmidt U. What does uncertainty mean to women with anorexia nervosa? An interpretative phenomenological analysis. Eur Eat Disord Rev. 2011;19(1):12–24. https://doi.org/10.1002/erv.1029.

    Article  PubMed  Google Scholar 

  51. 51.

    Konstantellou A, Hale L, Sternheim L, Simic M, Eisler I. The experience of intolerance of uncertainty for young people with a restrictive eating disorder: a pilot study. Eat Weight Disord Stud Anorexia Bulimia Obesity. 2019;24(3):533–40.

    Article  Google Scholar 

  52. 52.

    Sternheim L, Harrison A. The acceptability, feasibility and possible benefits of a group-based intervention targeting intolerance of uncertainty in adolescent inpatients with anorexia nervosa. Cogn Psychol. 2018;5(1):1441594. https://doi.org/10.1080/23311908.2018.1441594.

    Article  Google Scholar 

  53. 53.

    Vrieze E, Pizzagalli DA, Demyttenaere K, Hompes T, Sienaert P, de Boer P, et al. Reduced reward learning predicts outcome in major depressive disorder. Biol Psychiatry. 2013;73(7):639–45. https://doi.org/10.1016/j.biopsych.2012.10.014.

    Article  PubMed  Google Scholar 

  54. 54.

    Shields GS, Sazma MA, Yonelinas AP. The effects of acute stress on core executive functions: a meta-analysis and comparison with cortisol. Neurosci Biobehav Rev. 2016;68:651–68. https://doi.org/10.1016/j.neubiorev.2016.06.038.

    Article  PubMed  PubMed Central  Google Scholar 

  55. 55.

    Hernaus D, Quaedflieg CW, Offermann JS, Casales Santa MM, van Amelsvoort T. Neuroendocrine stress responses predict catecholamine-dependent working memory-related dorsolateral prefrontal cortex activity. Soc Cogn Affect Neurosci. 2018;13(1):114–23. https://doi.org/10.1093/scan/nsx122.

    Article  PubMed  Google Scholar 

  56. 56.

    Fairburn CG. Cognitive behavior therapy and eating disorders. Guilford Press; 2008.

  57. 57.

    Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59(20):22–33.

    PubMed  Google Scholar 

  58. 58.

    Overbeek I, Schruers K, Griez E. Mini international neuropsychiatric interview: Nederlandse versie 5.0. 0. DSM-IV [Dutch version]. Maastricht: Universiteit Maastricht; 1999.

    Google Scholar 

  59. 59.

    Shaffer D, Fisher P, Lucas CP, Dulcan MK, Schwab-Stone ME. NIMH Diagnostic Interview Schedule for Children Version IV (NIMH DISC-IV): description, differences from previous versions, and reliability of some common diagnoses. J Am Acad Child Adolesc Psychiatry. 2000;39(1):28–38. https://doi.org/10.1097/00004583-200001000-00014.

    Article  PubMed  Google Scholar 

  60. 60.

    Kovacs M. Children's depression inventory. North Tonawanda: Multi-Health System; 1992.

    Google Scholar 

  61. 61.

    Timbremont B, Braet C. Psychometrische evaluatie van de Nederlandstalige children's depression inventory. Gedragstherapie. 2001;34:229–42.

  62. 62.

    Buhr K, Dugas MJ. The intolerance of uncertainty scale: psychometric properties of the English version. Behav Res Ther. 2002;40(8):931–45. https://doi.org/10.1016/S0005-7967(01)00092-4.

    Article  PubMed  Google Scholar 

  63. 63.

    de Bruin GO, Rassin E, van der Heiden C, Muris P. Psychometric properties of a Dutch version of the intolerance of uncertainty scale. Neth J Psychol. 2006;62(2):87–92. https://doi.org/10.1007/BF03061055.

    Article  Google Scholar 

  64. 64.

    Rosser BA. Intolerance of uncertainty as a transdiagnostic mechanism of psychological difficulties: a systematic review of evidence pertaining to causality and temporal precedence. Cog Ther Res. 2019;43(2):438–63. https://doi.org/10.1007/s10608-018-9964-z.

    Article  Google Scholar 

  65. 65.

    Ashby FG, Gott RE. Decision rules in the perception and categorization of multidimensional stimuli. J Exp Psychol Learn Mem Cogn. 1988;14:33.

    Article  Google Scholar 

  66. 66.

    Brainard DH. The psychophysics toolbox. Spat Vis. 1997;10(4):433–6. https://doi.org/10.1163/156856897X00357.

    Article  PubMed  Google Scholar 

  67. 67.

    Maddox WT, Ashby FG. Dissociating explicit and procedural-learning based systems of perceptual category learning. Behav Process. 2004;66(3):309–32. https://doi.org/10.1016/j.beproc.2004.03.011.

    Article  Google Scholar 

  68. 68.

    Maddox WT, Filoteo JV. Striatal contributions to category learning: quantitative modeling of simple linear and complex nonlinear rule learning in patients with Parkinson's disease. J Int Neuropsychol Soc. 2001;7(6):710–27. https://doi.org/10.1017/S1355617701766076.

    Article  PubMed  Google Scholar 

  69. 69.

    Cohen J. Statistical power analysis for the behavioral sciencies. UK: Routledge; 1977.

    Google Scholar 

  70. 70.

    Maddox WT, Ashby FG, Bohil CJ. Delayed feedback effects on rule-based and information-integration category learning. J Exp Psychol Learn Mem Cogn. 2003;29(4):650.

    Article  Google Scholar 

  71. 71.

    Zeithamova D, Maddox WT. Dual-task interference in perceptual category learning. Memory Cog. 2006;34(2):387–98. https://doi.org/10.3758/BF03193416.

    Article  Google Scholar 

  72. 72.

    Bartlett MS, Kendall DG. The statistical analysis of variance-heterogeneity and the logarithmic transformation. Suppl J R Stat Soc. 1946;8(1):128–38. https://doi.org/10.2307/2983618.

    Article  Google Scholar 

  73. 73.

    Trafimow D, Wang T, Wang C, Myuz HA. Does slight skewness matter? Int J Aviation Res. 2019;11(1):11–24.

  74. 74.

    Olivo G, Gaudio S, Schiöth HB. Brain and cognitive development in adolescents with anorexia nervosa: a systematic review of FMRI studies. Nutrients. 2019;11(8):1907. https://doi.org/10.3390/nu11081907.

    Article  PubMed Central  Google Scholar 

  75. 75.

    Schilder C, Sternheim LC, Aarts E, van Elburg AA & Danner UN. Do adolescents with eating disorders ask too much of themselves in school? Relationships between educational achievement, intelligence and perfectionism. n.d.. under review IJED-20-0296.

  76. 76.

    Chang HJ, Kuo CC. Overexcitabilities: empirical studies and application. Learn Individ Differ. 2013;23:53–63.

    Article  Google Scholar 

  77. 77.

    Karpinski RI, Kolb AM, Tetreault NA, Borowski TB. High intelligence: a risk factor for psychological and physiological overexcitabilities. Intel. 2018;66:8–23.

    Google Scholar 

  78. 78.

    Jappe LM, Frank GK, Shott ME, Rollin MD, Pryor T, Hagman JO, et al. Heightened sensitivity to reward and punishment in anorexia nervosa. Int J Eat Disord. 2011;44(4):317–24.

    Article  Google Scholar 

  79. 79.

    Johnson JG, Cohen P, Kasen S, Brook JS. Eating disorders during adolescence and the risk for physical and mental disorders during early adulthood. Arch Gen Psychiatry. 2002;59(6):545–52. https://doi.org/10.1001/archpsyc.59.6.545.

    Article  PubMed  Google Scholar 

  80. 80.

    Kingston K, Szmukler G, Andrewes D, Tress B, Desmond P. Neuropsychological and structural brain changes in anorexia nervosa before and after refeeding. Psychol Med. 1996;26(1):15–28. https://doi.org/10.1017/S0033291700033687.

    Article  PubMed  Google Scholar 

  81. 81.

    Reville MC, O’Connor L, Frampton I. Literature review of cognitive neuroscience and anorexia nervosa. Curr Psychiatry Rep. 2016;18(2):18. https://doi.org/10.1007/s11920-015-0651-4.

    Article  PubMed  Google Scholar 

  82. 82.

    Tchanturia K, Anderluh MB, Morris RG, Rabe-Hesketh S, Collier DA, Sanchez P, et al. Cognitive flexibility in anorexia nervosa and bulimia nervosa. J Int Neuropsychol Soc. 2004;10(4):513–20. https://doi.org/10.1017/S1355617704104086.

    Article  PubMed  Google Scholar 

  83. 83.

    Lozano-Serra E, Andrés-Perpiña S, Lázaro-García L, Castro-Fornieles J. Adolescent anorexia nervosa: cognitive performance after weight recovery. J Psychosom Res. 2014;76(1):6–11. https://doi.org/10.1016/j.jpsychores.2013.10.009.

    Article  PubMed  Google Scholar 

  84. 84.

    Katzman DK, Christensen B, Young AR, Zipursky RB. Starving the brain: structural abnormalities and cognitive impairment in adolescents with anorexia nervosa. In Sem Clin Neuropsychiatry. 2001;6(2):146–52. https://doi.org/10.1053/scnp.2001.22263.

    Article  Google Scholar 

  85. 85.

    Sternheim L, Danner U, van Elburg A, Harrison A. Do anxiety, depression, and intolerance of uncertainty contribute to social problem solving in adult women with anorexia nervosa?. Brain Behav. 2020;10(6):e01588.

  86. 86.

    Roberts ME, Tchanturia K, Treasure JL. Is attention to detail a similarly strong candidate endophenotype for anorexia nervosa and bulimia nervosa? World J Bio Psychiatry. 2013;14(6):452–63. https://doi.org/10.3109/15622975.2011.639804.

    Article  Google Scholar 

Download references

Acknowledgements

Not applicable.

Funding

Not applicable.

Author information

Affiliations

Authors

Contributions

LS coordinated the study, obtained ethical approval, analyzed and interpreted the data and was a major contributor in writing the manuscript. MW collected the data for the Dutch samples, analyzed the data and contributed to writing the manuscript. UD coordinated the data collection for the AN Dutch sample and contributed to writing the manuscript.TM analyzed the data. VF analyzed the data. MS coordinated the data collection for the USA samples and helped interpreting the data. GF supervised LS, contributed to analyzing and interpreting the data and contributed to writing the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Lot C. Sternheim.

Ethics declarations

Ethics approval and consent to participate

USA: The study was approved by the Colorado Multiple Institutional Review Board. We obtained written assent from the participant and consent from the legal guardian prior to conducting any of the study procedures.

NL: This study was approved by the Medische Ethische Toetsingscommisie Universitair Medisch Centrum Utrecht (13–139). All participants and their parents or legal guardians gave consent for participation in this study.

Consent for publication

All participants and their parents or legal guardians gave consent for publication of the data acquired in this study.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Sternheim, L.C., Wickham, M.I., Danner, U.N. et al. Understanding implicit and explicit learning in adolescents with and without anorexia nervosa. J Eat Disord 9, 77 (2021). https://doi.org/10.1186/s40337-021-00431-7

Download citation

Keywords

  • Anorexia nervosa
  • Adolescents
  • Implicit learning
  • Explicit learning