From this initial dataset we kept only samples where the dating was based on Accelerator Mass Spectrometry (AMS)

We collected over 2,500 14 C dates from nine different prefectures: Tokyo, Kanagawa, Saitama, Chiba, Ibaragi, Tochigi, and Gunma (aggregated as Kanto region), Aomori, and Hokkaido (Fig 1). We also removed 14 C dates based on ples (i.e. shell and fishbone), as well as charred remains with high ? 13 C that might be potentially affected by a reservoir effect (i.e. >-24‰, given a range –24.0 ± 0.7 for terrestrial mammals and –25.9 ± 2.0 for C3 plants presented in ; a more conservative approach with a threshold set at -26‰ did not show major qualitative differences in the results, see S1 File). We excluded sites dated to Incipient and Initial Jomon periods (when the degree of sedentism was significantly lower), and those associated with the Final Jomon (when the overall sample size was too small in the Kanto region). We thus limited our analysis between the Early and the Late Jomon period, excluding all 14 C dates outside the interval 7,500

2,500 14 C years BP. The final sample size (n = 1,433; see Fig 1 for site distribution) used for the analysis was 406 14 C dates from 47 sites in the Kanto region, 432 14 C dates from 58 sites in Aomori Prefecture, and 595 14 C dates from 82 sites in Hokkaido. In order to reduce the effect of “ascertainment” or “wealth” bias we binned the 14 C dates based on the clustering of the mean 14 C years BP, using a threshold of 200 years (cf. ). Data, source codes and scripts used in this work can be found in S2 File and in the zenodo online repository (doi: /zenodo.47339).


The statistical comparison of the observed SPD of each study area against their respective fitted uniform and exponential null models (Fig 2) yielded a statistically significant global p-value (Table 1), indicating that some of the observed fluctuations are genuine, and not the result of sampling error, taphonomic loss, or a gradual exponential population increase. Local deviations from the null models highlighted Indonesian Cupid in all regions a general rise and fall pattern, with a steady growth starting around the beginning of the 7 th millennium BP, and general decline peaking towards the mid 4 th millennium BP (Fig 2). In the Kanto region, the population drop during the transition from the Early to Middle Jomon period is confirmed in the statistical comparison with the uniform model, with significant negative deviations around 5,700 cal BP. When compared against the exponential model, this decline is no longer significant, but the overall gradual decline starting after 5,000 cal BP leads the density of 14 C dates below the expectations of the null model at 3,900 cal BP first, and subsequently, after a temporary recovery, at 3,300 cal BP.

The thick lines show the 200-years rolling mean, whilst the grey band represents the 95% confidence interval for the null model (upper row: uniform distribution; lower row: exponential distribution). Red and blue vertical bands represent intervals with significant positive and negative deviations.

The resulting number of bins was 87 for Kanto, 128 for Aomori Prefecture, and 186 for Hokkaido

In Aomori Prefecture, the overall trajectory delineated by the SPD resembles a logistic curve, with the density of 14 C dates reaching an upper threshold at 5,500 cal BP. The subsequent interval between 5,500 and 4,000 seems to be relatively stable, with several minor fluctuations deviating from the null uniform model. When compared against the exponential model, the initial moments of this high-density stage is highlighted as significant positive deviation, as well as the temporary (and relatively sharp) drop observed around 3,800–3,700 cal BP.