Observing that an exposure (e.g. higher meat intake) is associated with an outcome (e.g. higher BMI) does not necessarily mean that the exposure caused the outcome. Other factors may "confound" the association by causing both the exposure and the outcome. Such confounding can be difficult to detect if the factors responsible have not been measured. A recent study proposed a method to detect confounding by unmeasured variables if they cause discontinuous variation in the exposure. We intend to develop this method further and apply it to the question of whether eating meat affects a person's BMI. The method should tell us whether simple observation of people's meat intake and BMI reveals the causal effect or is confounded.