Background: Earlier scales on mindful eating do not measure mindful eating independent from emotional or external eating, or mindful eating in common situations. Objective: The objective was to develop a new instrument to measure the attention element of mindful eating, the Mindful Eating Behavior Scale (MEBS), and to compute the internal structure, reliabilities, and convergent validity of this scale. Design: A cross-sectional ancillary study within the Longitudinal Aging Study Amsterdam was conducted between fall 2014 and spring 2015. Participants/setting: Participants were 1,227 Dutch adults aged 55 years and older from the Longitudinal Aging Study Amsterdam. Main outcome measure: A selection of 20 items from existing instruments was used to design an initial version of the MEBS. Statistical analyses performed: The internal structure of the MEBS was evaluated using an exploratory structural equation modeling approach on half of the sample and confirmatory factor analysis on the whole sample to develop the final version of the scale. The measurement invariance of the scores was tested with respect to sex, age, and body mass index. Reliabilities of subscales were determined with Cronbach's α. To test convergent validity, the scores of the new scale were correlated with theoretically relevant variables. Results: Two items were deleted because of low item loadings and one item because of high correlated uniqueness. The final confirmatory factor analysis model with 17 items and four domains (Focused Eating, Hunger and Satiety Cues, Eating with Awareness, and Eating without Distraction) showed good fit (comparative fit index=0.97, Tucker-Lewis index=0.96, and root mean square error of approximation=0.04). Measurement invariance was found for sex, age, and body mass index. Cronbach's α values were medium to high (.70 to.89). Most correlations were in the expected directions, which indicated good preliminary convergent validity. Conclusions: The MEBS was successfully developed consisting of 17 items and four domains. Because of low interfactor correlations, a total score combining the four domains should not be computed. The MEBS showed good internal consistency and preliminary convergent validity in a sample of Dutch adults aged 55 years and older.