Human and nonhuman faunal remains from small deposits at three sites, NGRP 16, 23, and 46, were analyzed to determine the species and minimum numbers of individuals represented, and to investigate the processes of natural and human-induced modification that resulted in the extensive fragmentation of the human remains. Taphonomic analysis indicates perimortem modification of the human remains during disarticulation and processing as well as pig predation. The types and locations of tool marks, fractures, and fracture products are quantified and described, and quantitative taphonomic profiles for the human remains are compared to human and nonhuman bone assemblages from middens and normative burial contexts on Fiji. The NGRP assemblages do not taphonomically match other assemblages believed to represent incidents of cannibalism, but this does not rule out the consumption of human flesh or decomposition fluids. We reject the notion of a universal taphonomic signature of cannibalism, and find it more constructive to examine these data in the context of archaeologically and ethnographically documented mortuary practice in the region. We propose that the assemblages from these sites represent secondary deposits of human remains that were part of multistage mortuary programs that included the curation of specific skeletal elements, as documented in ethnographic accounts of mortuary ritual across Melanesia and far into antiquity. This study presented an interesting test of how we identify and interpret evidence of mortuary behavior when we encounter assemblages of human remains that do not fit the traditional concept of a burial or mortuary feature as most archaeologists know it.