While antipredator strategies have been a focus of behavioral research for decades, scientists generally study the responses of prey toward overt, explicit threat. However, risk can also be significant when a threat is covert, such as when an ambush predator may be nearby or a secondary threat remains after a predator's departure. Little is known about the mechanism that prey use to assess risk in a predator's absence. Tests were conducted to determine the manner in which coyotes respond to these ambiguous threats. Specifically, I tested whether coyotes respond to prior anthropogenic activity that has occurred near their only food source, whether they investigate human activity at both profitable (feeding) and unprofitable (non-feeding) locations, and what sort of information coyotes are capable of gaining through their investigation. I explored these questions in three experiments spanning 4 years at the USDA/APHIS/WS National Wildlife Research Center's Logan field station. Test subjects were eight pairs of captive coyotes. Results showed that coyotes delayed or avoided feeding in response to prior anthropogenic activity, and that often a delay was due to investigation of human scent trails. Investigation of non-feeding areas occurred but was relatively brief. When coyotes were prevented from investigating locations of prior anthropogenic activity, foraging ceased altogether. In addition, coyotes were able to differentiate among the activity of different humans based on their association with negative, neutral, or positive threat levels, even in the presence of confounding visual and olfactory cues. They remembered these associations even after one month. This study is the first that provides evidence suggesting that canids gather and interpret complex information for cognitive inference about threat level associated with access to food.