Description: Objective Diarrhoea is a significant contributer to morbidity and is among the leading causes of death of children living in poverty. As such, the incidence, duration and severity of diarrhoeal episodes in the household are often key variables of interest in a variety of community-based studies. However, there currently exists no means of defining diarrhoeal severity that are (A) specifically designed and adapted for community-based studies, (B) associated with poorer child outcomes and (C) agreed on by the majority of researchers. Clinical severity scores do exist and are used in healthcare settings, but these tend to focus on relatively moderate-to-severe dehydrating and dysenteric disease, require trained observation of the child and, given the variability of access and utilisation of healthcare, fail to sufficiently describe the spectrum of disease in the community setting. Design Longitudinal cohort study. Setting Santa Clara de Nanay, a rural community in the Northern Peruvian Amazon. Participants 442 infants and children 0–72 months of age. Main outcome measures Change in weight over 1-month intervals and change in length/height over 9-month intervals. Results Diarrhoeal episodes with symptoms of fever, anorexia, vomiting, greater number of liquid stools per day and greater number of total stools per day were associated with poorer weight gain compared with episodes without these symptoms. An instrument to measure the severity was constructed based on the duration of these symptoms over the course of a diarrhoeal episode. Conclusions In order to address limitations of existing diarrhoeal severity scores in the context of community-based studies, we propose an instrument comprised of diarrhoea-associated symptoms easily measured by community health workers and based on the association of these symptoms with poorer child growth. This instrument can be used to test the impact of interventions on the burden of diarrhoeal disease.