Jomon Population dynamics as inferred by site and residential unit counts

While site-counts were, and still are, used as proxies, several authors also offered time-series of pit-dwellings [7,16–17], arguing that these provide a more dependable alternative as well as the possibility to rely on a more refined pottery-based chronology, occasionally offering even sub-century temporal resolution

Subsequent studies have offered more detailed, yet fragmented images of Jomon demography. Most have focused on smaller regions, sacrificing the broad comparative dimension offered by Koyama’s work. In many cases, these studies have identified fluctuations occurring at scales that were not visible in Koyama’s analysis, prompting many to look at potential correlation with climate change [7–8,19].

Consequently, cross-regional comparisons are strongly affected by the proposed matching of different pottery-phase sequences, and the comparison with environmental and climatic data are hindered by chains of indirect dating

While these studies offer insights on Jomon population dynamics, their inferential power is to some degree limited by: 1) a substantial lack of statistical evaluation for distinguishing genuine fluctuations from sampling error; and 2) an almost exclusive reliance on a relative chronological framework, rather than absolute calendar dates. Given the large-scale investment in rescue archaeology and a long tradition of pottery typological studies, it is undeniable that the latter offers an economically viable option with robust sample sizes (n>1,000) and the possibility to rely on a fairly detailed sequence (i.e. more than 50 phases for ca 10,000 years). Yet, sample units (i.e. sites and/or residential features) are indirectly dated based on the recovery of diagnostic artefacts, and hence the assignment of the former to specific chronological blocks (periods or phases) is constrained by the quality and the quantity of the latter (see [20,21] for discussion). In many instances, raw data have mixed quality and resolutions and time-series generated from these do not adequately quantify the degree of temporal uncertainty (but see ). Furthermore, studies attempting to build an absolute chronological referencing for Jomon pottery phases are still rare and the few exceptions [18,22–24] do not formally address the question of the shape of the underlying distribution of 14 C dates associated with specific phases (cf. ), nor attempt to quantify the uncertainty of the retrieved parameters (as in [26–27]).

This paper seeks to overcome some of these limitations by using summed probability distribution (SPD) of 14 C dates to infer Jomon population dynamics. We argue that while counts of pit-dwellings might offer a more direct proxy, the limits imposed by its reliance on a relative chronological framework urge the exploration of alternative proxies. We analyse the SPDs of three regions from Eastern Japan (Hokkaido, Aomori Prefecture, and Kanto; see Fig 1) that were densely populated during the Jomon period: to determine whether key fluctuations are genuine or the result of sampling error and to identify possible divergences in the population trajectory of the three areas within an absolute chronological framework. In order to achieve these objectives, we adopt a hypothesis-testing approach for assessing the SPDs of 14 C dates, introducing a new permutation-based technique apt to statistically compare different datasets.

As mentioned, the temporal framework of Jomon archaeology is rarely based on absolute chronology and in most cases it relies on pottery-phases or broader periodisations. Here we briefly review existing attempts of population reconstruction in the three study areas examined in this paper, with reference to such relative temporal frameworks. Although a conclusive agreement is not established, Kobayashi has examined 14 C dates associated with many of these pottery phases as well as all major Jomon periods, and proposed an absolute chronological sequence. Unless otherwise stated, in the paper we use this chronological reference. In particular, we focus on the interval between 7,000 and 3,000 cal BP, which approximately corresponds to the Early (6,950