Algonquian languages show a hierarchy effect in which Infl preferentially agrees with the argument that ranks highest on the hierarchy SAP > proximate 3 > obviative 3. When this argument is the object, the outcome is a special agreement pattern known as the “inverse”. The Algonquian inverse is well-known, but what has gone unnoticed in the theoretical literature is that two Plains Algonquian languages, Arapaho and Gros Ventre, have developed a second, entirely distinct hierarchy effect in which Voice preferentially agrees with a first-person nominal and Infl concurrently agrees with the leftover non-first-person nominal. The two hierarchy effects coexist within the same paradigm: some forms show the familiar inverse pattern driven by Infl; others show the distinct “speaker-centric” pattern driven by Voice. In contexts that could conceivably trigger either pattern, the Voice-driven pattern bleeds the Infl-driven pattern. I argue that the existence of the two patterns, their distinct effects on the overall verb form, and their distribution in paradigms follows from the existence of relativized probes on both Voice and Infl.