When compared against a uniform model, all three regions show most episodes of negative divergence before c 5,000 cal BP, and most of the positive divergences afterwards. When compared against exponential models the similarities between the three regions are even more striking. In all cases we see positive deviations between 6000 and 5,000 cal BP and a negative deviation after 3,700 cal BP. The results are thus consistent with a general rise-and-fall pattern, where the density of 14 C dates after the peak is higher than what is observed before.
The x-axes are in cal BP
Despite these broad similarities, our analyses indicate the presence of some local divergence in the SPDs. Fig 3 and Table 1 shows the output of the non-parametric pair-wise permutation tests. The global p-values (Table 1) are mostly non-significant, albeit the comparison between Aomori and Kanto returned a p-value of 0.0555. The local analysis (Fig 3) highlights portions of the SPDs where we observe significant divergences. While none of the divergences between the SPDs of Aomori and Hokkaido can be statistically supported (the global p-values are both >0.9), both regions exhibit some significant differences with Kanto. Aomori shows higher density in 14 C dates around 5,700 cal BP (corresponding to the decline observed in Kanto between the end of Early Jomon and the beginning of Middle Jomon), and a drop in density around 5,200 cal BP (when the SPD in Kanto is showing a steady growth). Although the global p-value of Hokkaido against Kanto is non-significant (p = 0.2752), we observe local positive deviations again at 5,700 cal BP, and a negative deviation at 5,000 cal BP, corresponding to maximum peak of 14 C dates in Kanto. Given that all local deviations were in the first 2,000 years, and that the SPDs exhibit substantial similarity during the last 1,500 years, we executed the global significance test reducing the temporal scope to the Early and Middle Jomon periods (7,000
4,420 cal BP, ; Table 1). This led to a highly significant divergence between Aomori and Kanto (p = 0.0039), but not between Hokkaido and Kanto (p = 0.1124). In all cases when the focal set was Kanto the p-values increased (p = 0.1227 against Aomori and 0.667 against Hokkaido), most likely due to the smaller sample size of this study area.
Each row represents an observed SPD of a region compared against another in the column. Red and blue vertical bands represent intervals with significant positive and negative deviations from a null model of the aggregated set of each of pair.
Discussion
Our study provides the first application of SPD analysis for Jomon data, allowing a formal re-assessment of its population dynamics as suggested by the density of 14 C dates. The case studies offered an exceptionally high density of samples, with a total sample size (n = 1,433) comparable to much larger, continental scale study areas (cf. [39,44]). In qualitative terms, the results confirm some of the general trends suggested by previous works [7,17,21,57] but, more importantly, offers an absolute chronological framework for establishing the timing of key events. Regrettably, the high degree of chronometric uncertainty for sites and pithouses makes the statistical comparison with SPDs unfeasible, as genuine similarities or dissimilarities cannot be distinguished by spurious patterns determined by calibration wiggles and sampling error. We however review our results in relation to existing studies as this can still offer some insights to Jomon population dynamics, but we stress the fact that these comparisons are limited to broad scale qualitative accounts be naughty.
In the Kanto region, our analysis indicates a significantly lower density of 14 C dates around 5700 cal BP compared to the other two regions and the uniform null model. This corresponds to the decline in the counts of residential units observed between the end of the Early Jomon and the beginning of the Middle Jomon (corresponding to the interval between the Moroiso c and the Goryogadai 2 phases [7,58–59]; approximately dated to 5,700
