In striking contrast to the voice-marking affixes, for which there is universal agreement, the case-markers of Proto-Austronesian (PAn ), and later protolanguages such as Proto-Philippines (PPh ), have long resisted detailed reconstruction. Four sets of case-markers (nominative, genitive, oblique, locative) can be posited for PAn , each distinguished by its application to personal names vs. common nouns, and singular vs. plural reference in the former category. Through standard use of the comparative method, it is shown that PAn must have had at least the genitives *ni, *na, *nu, and nominatives *si, *sa, where *i correlates with singular personal names, *a with plural personal names, and *u with common nouns. For PPh , application of the comparative method allows the nominatives to be expanded to *si, *sa, *su. Although the oblique and locative sets show much greater functional variability, the case-markers *ki, *ka, *ku, and *di, *da must be reconstructed for PAn (plus *du for PPh ), thus presenting the analyst with a glossing challenge for these forms that is not found with *ni, *na, *nu, or *si, *sa, (*su). By a process of elimination, *ki and *ku can be posited for PAn , Proto–Malayo-Polynesian (PMP), and PPh as the oblique of singular personal names and of common nouns, respectively. Although PPh *ka can be added to this set as the oblique of plural personal names, this interpretation can be applied to PMP and PAn only by generalizing the semantic values of the *i/a/u vowel template to cases where they are not directly attested. For the locatives, *di can be reconstructed without further specification, but *da and *du present additional challenges that are ultimately met by allowing the boundary between obliques and locatives to be permeable. The result is four tripartite sets of case-markers for PPh (nominatives *si, *sa, *su, genitives *ni, *na, *nu, obliques *ki, *ka, *ku, locatives *di, *da, *du), of which all but *su and *du can be attributed to PAn , and all but *du to PMP, suggesting replacement of these terms in all Formosan languages. A fifth set of case-markers *i, *a, *u may have existed, but details regarding its function remain elusive.