Objectives: The influence of food intake on behavioural disorders was already described in the early 20th century. Elimination of individually allergenic food items from individual diets ["oligoantigenic diet" (OD)] showed promise to improve attention-deficit/ hyperactivity disorder (ADHD) symptoms. However, only few of the positive results were evaluated by blinded symptom rating. Therefore the present study's purpose was to evaluate the reliability of a non-blinded rating of the ADHD Rating Scale IV (ARS) for the assessment of OD effects in comparison to a blinded rating of the ARS based on pseudonymized video recordings. Methods: Ten children (8m/2f) aged 8 to 14 with ADHD according to ICD-10 participated in an uncontrolled, open-label dietary intervention study. Food items, commonly related to intolerances, were eliminated for four weeks. Participants with > 40% improvement in the ARS between T1 (before the diet) and T2 (after the diet) were defined as responders. Nutrients with individual relevance to ADHD symptoms were identified in a following reintroduction phase (T3-T4) lasting 8-16 weeks. The ARS was completed by a nonblinded child and adolescent psychiatrist (T0-T4). Sessions were recorded on video, pseudonymized, and evaluated by three blinded raters. Complete data were captured for eight children. The inter-rater reliability between the non-blinded therapist and every blinded rater was determined by the intra-class correlation coefficient (ICC). Correlations according to Pearson and Spearman between the non-blinded and blinded rating were calculated for each rater. Results: Two blinded raters and the non-blinded rater considered 5 of 8 (62.5%) children as responders, whereas one blinded rater disagreed as to the success of one case thus considering only 4 of 8 children as responders to the diet. Inter-rater reliability was assessed after each rater having scored 33 videos: The intra-class coefficients were >.9 for all raters (rater 1: ICC=.997, rater 2: ICC=.996, rater 3: ICC=.996) and the Spearman rho between the raters were high (n=33; rater 1: rho =.989, p<.0001, rater 2: rho=.987, p<.0001, rater 3: rho=.984, p<.0001), respectively.