This paper investigates the effect of drinking arsenic contaminated water on mental health. Drinking water with an unsafe arsenic level for a prolonged period can lead to arsenicosis and associated illness. Based on rich and newly collected household survey data from Bangladesh, we construct several measures for arsenic contamination that include the actual arsenic level in the respondent’s tubewell (TW), and past institutional arsenic test results as well as their physical and mental health. In contrast to the existing literature, we provide objective measures of arsenic exposure and take advantage of the quasi-randomness of arsenic distribution to account for the potential endogeneity of developing arsenicosis related to possible selection of certain households into using safe or unsafe sources of water. We take the pre-1999 use of TW as an instrument and structural modelling as alternatives for robustness checks. We find that suffering from an arsenicosis symptom is strongly negatively related to mental health, even more so than from other illnesses. While we cannot disentangle the specific mechanisms that drive the results, we do provide a framework for thinking about the role that physiological, social, and psychological factors may play on it.