The empirical robustness of emotional eating and stress eating - the tendency to eat in response to affective experience - is under ongoing debate. Is this a general phenomenon or is this seen only in certain groups? And, if so, can such interindividual differences be reliably assessed with self-report questionnaires? Here, we approach this question not through measuring overt eating - a behavior that is subject to several contextual conditions - but through food craving, an important precursor to food intake that can be triggered rapidly and involuntarily. We also study temporal ordering, i.e., whether affect cooccurs with, precedes, or follows craving. We pooled 8 ecological momentary assessment (EMA) studies comprising 764 participants answering 4-6 daily questionnaires for 7-20 days (mega-analysis). Trait-level eating style questionnaire scores were modeled as moderators for potential emotion/stress-craving relationships. Negatively toned affect (feeling bored, irritated, stressed) but also positive affect (cheerful, enthusiastic, relaxation with reverse direction) co-occurred with more craving. Feeling bored, stressed, or less relaxed, calm, and cheerful was also associated with later craving which, in turn, related to feeling more relaxed and less stressed at the next time point. Eating style questionnaires moderated concurrent but not prospective affect-craving or craving-affect relationships. Our results highlight the disparities between questionnaires and EMA measures, particularly for prospective relationships. Affect and craving seem to interact in close temporal proximity in this mostly healthy weight sample whereas the predictive validity of eating style questionnaires for naturalistic contexts is limited. We suggest a refined terminology to reflect these findings.