Millions of cotton farmers in India use Monocrotophos, a pesticide that is both toxic—the level of exposure found in the field is linked to irreversible cognitive impairment, depression and suicidal tendencies—and inferior in efficacy to other safer, similar cost alternatives. I use three experiments to test whether misperception and inferential challenges created by a habit of mixing different inputs together explain why farmers fail to abandon Monocrotophos. I conduct a brief information campaign that reduces farmers’ self-reported plans to purchase the pesticide for the next planting season by 37%. I show how the campaign addresses behavioral biases. I discuss implications for public health policy interventions and general lessons for thinking about mechanisms for technology selection in markets where “unlearning” specious product benefits may be difficult due to psychological and behavioral stumbling blocks.