This paper examines the causal relationship between climate shocks and collective violence in India using annual data over the period 1954–2006. We use the ARDL bounds testing approach to deal with problems of autocorrelation and non-stationarity of key variables. Rather than rainfall, we find that it is maximum temperature that has long and short run effects on collective violence, with unidirectional causality from temperature shocks to riots. A one standard deviation increase in maximum temperature over the long run average increases the number of riots by 55 per cent. Return to long run equilibrium after a temperature shock takes approximately 15 years. The insignificance of rainfall holds whether we consider rainfall levels or rainfall growth. Given the absence of long run relationships between income levels /growth and riots, it is unlikely that the income channel is the one through which climate affects riots in India. Instead, the evidence suggests a psychological channel through which temperature affects riots in India, with hotter temperatures being associated with increased levels of aggression.