Participants (Table 1). The present sample included 27 women with a diagnosis of current BN, 28women with a diagnosis of current AN, and 27 healthy womenwithout any current or lifetime Axis I diagnosis.Individuals with AN were recruited from the inpatient unit and individuals with BN from the outpatient unit at the Center for Eating Disorders, Department of LiaisonPsychiatry and Psychosomatic Medicine, University Hospital of Zurich, Switzerland. Healthy control participants were recruited from local universities, colleges and vocational schools using ﬂyers and electronic advertisements. Participants were female and at least 18 years of age (M = 22.6, range 21.4-24) and without past or current neurological disorders or professional knowledge about nutrition (e.g., cook or dietician).Cognitive impairment was ruled out using the Trail Making Test during the study.Data from HC and individuals with AN were previously compared in Milos et al. . For more information on exclusion/inclusion criteria of AN and HC participants, refer to Milos et al. .
Standardized diagnostic interview. The presence of past and current psychiatric disorders, including ED, was determined with the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) .
Standardized clinical questionnaires. The severity of depressive symptoms was assessed using the Beck Depression Inventory (BDI)[11, 12, 13]and eating disorder symptoms and features were assessed using the Eating Disorder Inventory (EDI).Besides yielding a total score, EDI also allows for the assessment of ED related symptoms such as drive for thinness, bulimia, body dissatisfaction, ineffectiveness, perfectionism, interpersonal distrust, interoceptive awareness, and maturity fears. Trait anxiety was measured by the State-Trait Anxiety Inventory (STAI) [15, 16].
Figure Rating Scale. First, subjects watched a video of a functionalized female body. When they clicked on "play", this body became thicker, on a scale from 01 to 80. They could pause and rewind the video at any time. First, they indicated how they saw their body at the moment and then how they would like to be[17, 18].
Hunger assessment. A one-item hunger scale  was administered to assess the current hunger state. When presented with a visual analogue scale ranging from not full/not hungry, to extremely full/satiated, participants wereasked to indicate their hunger level .
Trail Making Test.Cognitive performance was examined using the paper-and-pencil version of the Trail Making Test (TMT) for attention and set-shifting[21, 22]. This Test consists of two parts, each preceded by a short exercise. In the first part, participants were asked to draw lines connectingthe numbers from 1 to 25 as quickly as possible and without lifting the pencil from the paper. The beginning (1) and the end (25) are indicated. At this time, the timer is stopped, and the number of mistakes is counted. If a mistake (omitting a number or lifting a pencil) is found, the respondent is made aware of this and must continue from the last correctly marked number, while the timer continues to run. Time to complete the task and the number of errors were recorded. The second part had the same procedure with the difference that the numbers from 1 to 13 and the letters from A to L were on the sheet. Participants were then told to connect numbers with letters with numbers with letters and so on whereby both numbers and letters had to be in the correct order.That is, a line should be drawn from 1 to A, from A to 2, from 2 to B, from B to 3, and so on up to 13. Again, total time to complete the task and number of errors were recorded.
Portion estimation task. As in Milos et al. participants performed a computer-based task that required the size of sequentially displayed meal portions to be estimated on a visual analogue scale (0 = small, 100 = large). The meal pictures consisted of a breakfast meal (bread, butter, jam, yoghurt or muesli, orange juice), a main course A (meat, risotto, broccoli, fruit salad, with or without icecream) and a main course B (lasagna, salad, applesauce with vanilla cream). In main course A, macro nutriments of the meal were better recognizable than in main course B. Each meal was depicted by six different serving sizes: 1/8, ¼, ½, 1, 1 ½, 1 ¾ (pictures of the apportioned meals are available on Milos, et al., ). A single presentation of each of the 18 meal-portion pictures was incorporated into two blocked conditions (general, intent-to-eat). In the general condition, participants were simply instructed to estimate the size of the meal portion in general; in the intent-to-eat condition, participants were instructed to estimate the size of the meal portion while imagining they were supposed to eat the pictured meal. The sequences of pictures and conditions were randomly determined. After participants ﬁnished estimating the meal portion sizes, the full set of 18 meal pictures was displayed on the screen in a randomized arrangement (6 pictures from course A, then 6 pictures of course B, and then 6 pictures of breakfast) and participants were asked to order the depicted meals according to their size, beginning with the smallest one. Meal portion sizes were determined by the local nutritional advisor from the University Hospital of Zurich (Division of Endocrinology, Diabetes and Clinical Nutrition) according to guidelines of the Swiss Society for Nutrition. Carbohydrates, proteins and fat were ideally balanced in each meal. Based on this information and the assumption that 600–700 kcal represents a normal meal for 18-30 years old women, a normal meal portion (size 1) was deﬁned and used as the basis for deriving the other meal portion sizes.
Procedure. Participants were instructed to eat a meal 2 hours prior to the experiment, and refrain from eating after that, in order to minimize the effect of hunger on meal portion size estimates . Participants underwent the diagnostic interview, followed by the neuropsychological test, standardized clinical questionnaires and the experimental task, in that sequence. The entire procedure took approximately 3 hours.
Data analysis. T-tests comparing the clinical and cognitive assessmentsand exploratory correlation analyses were performed using R  (Version 1.3.1073).Mixed models of the portion estimation task were calculated using IBM SPSS Statistics 22 (IBM Corp., Armonk, NY, USA). To avoid retesting of previously reported contrasts, two separate models were calculated to compare theindividuals with BN to healthy controls (Model I) and to individuals with AN (Model II).These two models included Group(bulimia nervosa (BN)vs. healthy controls (HC) for Model I, bulimia nervosa (BN) vs. anorexia nervosa (AN) for Model II),Intent (general condition, intent-to-eat condition), Portion size (1/8, 1/4, 1/2, 1, 1 1/2, 1 3/4) as fixed factors. Subject was treated as a random intercept with a variance component covariance structure. The trial sequence was modelled as repeated measurements with a diagonal covariance structure.Bonferroni-Holm corrected pairwise comparisons of the estimated marginal means were used as post-hoc tests when applicable.
Exploratory analyses. We further explored the relationship of portion estimation and eating disorder symptoms across all participants. We created a single portion estimation item (a means of all estimation trials) and explored Pearson’s correlation analyses between this item and each subscale of the EDI, hunger assessment, STAI and BDI.