OverviewMachine learning (ML) is rapidly being adopted by archaeologists interested in analyzing a range of geospatial, material cultural, textual, natural, and artistic data. The algorithms are particularly suited toward rapid identification and classification of archaeological features and objects. The results of these new studies include identification of many new sites around the world and improved classification of large archaeological datasets. ML fits well with more traditional methods used in archaeological analysis, and it remains subject to both the benefits and difficulties of those approaches. Small datasets associated with archaeological work make ML vulnerable to hidden complexity, systemic bias, and high validation costs if not managed appropriately. ML's scalability, flexibility, and rapid development, however, make it an essential part of twenty-first-century archaeological practice. This review briefly describes what ML is, how it is being used in archaeology today, and where it might be used in the future for archaeological purposes.

Wilmshurst et al. (1) proposed recent and rapid colonization of East Polynesia based on analysis of 1,434 radiocarbon determinations. We commend the development of rigorous and replicable radiocarbon protocols that emphasize accuracy and precision, but we found (i) inaccuracies in their originally published supplementary data table, (ii) problems with their criteria for exclusion and inclusion of valid colonization estimators (i.e., Class 1 dates), and (iii) biases in their statistical analysis.Our review of their originally published 207 Class 1 dates identified 112 incorrectly reported 14 C laboratory numbers and 123 misreported conventional radiocarbon ages, with 110 of these reported as at least 100 y too recent. Additionally, source citations were misassigned for 70 Class 1 dates. Nonetheless, our reanalysis using corrected data provides probability distributions broadly similar to figure 4 in ref.1. The errors have been corrected in a revised table.We suggest that some reliability classification criteria for Class 1 dates are overly strict and exclude accurate estimators of early cultural activity. Specifically, several reliable dates (on archaeological criteria) with SEs of 10-15% are excluded by their 10% threshold, whereas elimination of all marine samples, even in cases where local ΔR values are established (2), seems inappropriate.Inclusion of dates as recent as 300 B.P. and samples from nonbasal strata biased their age estimation models in favor of a short chronology. These late dates skewed their sum of probability distributions to the more recent period, thereby affecting the cumulative probability outcomes. Fig. 1 shows the impact of removing late dates and how easily such probability curves can be affected by small sample sizes, which is the case for most archipelagoes.Also, cutoff points are assigned to identify the upper limit of likely colonization (e.g., 1300 A.D. for most islands), times by which Wilmshurst et al. (1) had "100% confidence that colonization had occurred" (1). These are based on the skewed probability sums (above), which influenced the slope of the cumulative probability line. Using only the earliest Class 1 dates (specific to each archipelago) results in different summed probability and cumulative curves as well as different colonization models (Fig. 2).Overall, we agree with Wilmshurst et al.(1) and others (3) that East Polynesia was settled more recently than previously argued. However, their statistical model was built on 14 C dates with calibrated probabilities that were summed, normalized, and then compared with a certainly settled date. We suggest that the analysis of probability distributions is more appropriate for identifying the timing of established settlement rather than initial colonization (4). Our reanalysis using their approach and a corrected version of their table S1 for Class 1 dates suggest that, in several cases, colonization probably occurred earlier than they proposed (1). We argue that several aspects of their reliability classification and statistical...

No abstract

The results of chemical and petrographic.

Archaeological survey on Muyuw (Woodlark Island) in the Massim area of Papua New Guinea located a number of stone arrangements, commonly known as megaliths. Test excavations have revealed the use of the stone arrangements as burial structures. The Muyuw data show a complex pattern of changing internal relationships and regional political relationships. The presence of stone arrangements in all the major islands of the northern Massim (and possibly beyond), hints at a shared regional symbolic system for dealing with the dead, and organising labour for public work. Stone arrangements form a complex Early Period (~1500BP–600BP) landscape built for the dead to negotiate relationships between the living throughout the northern Massim. Yet by 600 BP, this landscape had probably lost its symbolic potency. These sites are discussed in relation to the prehistory of the island and the region as a whole.