This paper argues for an extension of current models of Agree to capture relativized EPP effects, where a probe for movement targets an element with a specific set of features. We support the proposal through a case study of long distance agreement (LDA) in the Border Lakes dialect of Ojibwe (Central Algonquian), where the patterns of LDA depend on the particular combination of person/animacy features of the embedded arguments. This can be captured by the feeding and bleeding relationships between agreement and movement probes on Voice, Infl, and C.