Conclusion
The use of SPD for inferring past population dynamics has recently sparked debate, raising questions concerning the reliability of the underlying data and its general validity as a proxy (e.g. [45,74–76]). As it stands, the advantages provided by SPD over alternative methods outweighs its potentials limitations. We however stress that no proxy provides a unique and definitive portrait of prehistoric demography, and that when possible, multiple lines of evidence [70,77,78] should be assessed in comparative fashion. Furthermore, we would like to highlight that SPDs, as well as other count-based proxies, should never be simply assessed in visual terms, but compared against formal statistical models. Future studies should also go beyond the hypothesis-rejecting framework, and aim to statistically compare multiple working hypotheses. Recent advances in likelihood-free inferential methods such as Approximate Bayesian Computation [79–80] is offering potential new lines of research that has already been applied for the study of prehistoric demography .
The SPD of the 14 C dates from Hokkaido, Aomori, and the Kanto region do exhibit patterns that closely match other lines of evidence [14,57], offering at the same time an absolute chronology of key demographic events pertaining to the general rise-and-fall pattern observed in northeastern Japan. In particular, the number of 14 C dates exhibits a rapid exponential increase between 6,000 and 5,000 cal BP, followed by a high-density interval between 5,000 and 4,000 cal BP, a decline with a trough at c. 3,700 cal BP, and a renewed growth in the subsequent centuries, peaking at c. 3,300 cal BP. In broad terms, this pattern is consistent with the idea of a higher population density between the Bond 4 (5,900 cal BP) and 3 (4,200 cal BP) events , although further studies based on local climatic proxy will be necessary to establish this association (cf. [8,61]).
Albeit all three areas share broad similarity in their population history, our analysis also indicates that the SPD of the Kanto region show two significant divergences when compared to the other two areas: a temporary decline at c. 5,700 cal BP, and a peak at c. 5,000 cal BP immediately followed by a continuous decline till 3,700 cal BP. This is in contrast to a more stable “plateau” pattern during the 5 th millennium cal BP, and a decline starting only at 4,000 cal BP observed in the Hokkaido and Aomori regions. In general terms, these patterns support previous studies based on the time-series of pit-dwelling counts, but at the same time overcome many of the limitations imposed by a relative chronological framework. The SPD analysis’ major results are the ability to define the absolute timing of these events, as well as substantiating that these regional divergences are not due to simple mismatches in the relative pottery-based chronological sequences nor to sampling errors. Establishing how much of the observed differences in the SPDs are due to underlying divergences in the population history and how much to variation in subsistence-settlement strategies remains, however, an intriguing open question. A combination of multi-proxy demographic analyses with independent lines of evidence, a comparative assessment of the subsistence data, and formal evaluation of the temporal relationship with climatic changes are the next key directions to be undertaken in this regard.
Supporting Information
Time-series of pithouse and site counts: a) pithouse counts in Saitama, Kanagawa, and Tokyo prefectures (95% confidence intervals obtained from 1,000 Monte-Carlo iterations; details in ); b) rate of change estimates of pithouse counts in Saitama, Kanagawa, and Tokyo prefectures (95% confidence intervals obtained from 1,000 Monte-Carlo iterations; details in ); c) site counts in Hokkaido (data from ); d) site counts in Aomori (data from ); e) pithouse counts in Aomori (data from ); f) pithouse counts in the Hachinohe City area (data from [17,33]).
Given these premises, it is not surprising that Japanese archaeologists have long been trying to reconstruct Jomon population dynamics from the archaeological record in order to identify relationships with changes in the subsistence-settlement system. Early studies in the 1960s have identified major regional differences in the number of archaeological sites attributed to the Jomon culture, with the northeastern portion of the Japanese archipelago showing a considerably higher density than the southwest. This led scholars, such as Yamanouchi , to suggest that this pattern was reflecting the underlying heterogeneity in the distribution of key resources such as acorn, chestnut, and salmon (see also ). These early studies did not, however, seek to reconstruct temporal changes in the Jomon population size in a systematic manner, an endeavour that was subsequently pursued by Koyama [14–15]. His analysis confirmed and added more detail to these earliest impressions of spatial variability, providing at the same time a diachronic perspective on Jomon population change. Albeit based exclusively on sites counts and framed by a relatively coarse temporal (archaeological periods of ca 1,000 years) and spatial (regions between 30,000 and 80,000 km 2 ) resolutions, the scale of Koyama’s study is still unmatched, and its estimate of absolute population sizes remains the sole attempt proposed so far.
3,220 cal BP) Jomon periods. It is worth pointing out that although this Match.com scheme is used as a broad chronological reference for eastern Japan (see for example ), the 14 C dates underpinning Kobayashi’s study were primarily retrieved from Kanto, hence the synchronicity of the cultural phases of the three regions is not warranted [18, 28–29].
Materials
Finally, in Hokkaido the SPD highlights a steady, uninterrupted growth from c 6,000 cal BP to 5,000 cal BP, followed by a temporary small drop at ca. 4,400 cal BP (albeit not a significant deviation from either of the two models), a short recovery peaking at 4,000 cal BP (significant against both null models), and a more consistent decline with the lowest density of 14 C dates recorded around 3,700 cal BP (significant deviation against the exponential but not the uniform model).
5,380 cal BP ). While our SPD show some differences in the timing of the subsequent increase in the population trajectory when compared with previous works (at 5,500 cal BP in our study, later in [9,21]), the matching is quite remarkable, and confirms possible correlation with several environmental changes linked with the Bond 4 event (ca. 5.9 k cal BP ; see for a recent discussion). However, in Aomori Prefecture and Hokkaido, this period shows evidence of a steady growth, suggesting that a continental climate-based hypothesis should take into account both local environmental variations and potentially divergent cultural responses in different parts of the Japanese archipelago.