Abstract
Sea otters are well-known tool users, employing objects such as rocks or shells to break open invertebrate prey. We used a series of generalized linear mixed effect models to examine observational data on prey capture and tool use from 211 tagged individuals from 5 geographically defined study areas throughout the sea otter’s range in California. Our best supported model was able to explain 75% of the variation in the frequency of tool use by individual sea otters with only ecological and demographic variables. In one study area, where sea otter food resources were abundant, all individuals had similar diets focusing on preferred prey items and used tools at low to moderate frequencies (4–38% of prey captures). In the remaining areas, where sea otters were food-limited, individuals specialized on different subsets of the available prey and had a wider range of average tool-use frequency (0–98% of prey captures). The prevalence of difficult-to-access prey in individual diets was a major predictor of tool use and increased the likelihood of using tools on prey that were not difficult to access as well. Age, sex, and feeding habitat also contributed to the probability of tool use but to a smaller extent. We developed a conceptual model illustrating how food abundance, the prevalence of difficult-to-access prey, and individual diet specialization interacted to determine the likelihood that individual sea otters would use tools and considered the model’s relevance to other tool-using species.