Before the Leaving: Population Growth, Common Observance and Settlement Founding in the Last 400 Years of Eight Civilisations Recorded as Departing the Real, with Twelve Comparison Civilisations

1. Introduction

The Sublimed are the civilisations and individuals who have left the Real for a higher-dimensional existence, and they rarely interact with the Involved, the societies still active in the physical galaxy. What the Involved can learn of Sublimation by observation is therefore what observers saw before the leaving and at its end (Brennaquist, 1822 CE). This paper stays inside that boundary. It concerns the last centuries of civilisations before they departed, and it says nothing of what follows departure, which the records we use do not describe.

Involved scholars have proposed several signs that a civilisation is approaching departure. Drevoly (1856 CE) argued that societies converge on shared observances in their final centuries. Vandrasque and Lissome (1877 CE) described a slowing of population growth in three departing civilisations and treated it as a fertility transition. Quennevar (1869 CE) observed that Sublimation is recorded so seldom that any such sign rests on a handful of cases. None of these studies set departing civilisations beside civilisations that did not depart, and so none could say whether the signs were distinctive or ordinary features of long-lived societies.

We make that comparison using Contact's observation records. We write in 1900 CE, and both archive series we use were closed for this study in 1893 and 1895 CE; nothing after those years enters the paper. We ask whether, in the 400 years before departure, eight civilisations differed from twelve that did not depart in three respects: the growth of their populations, the spread of a common observance among their settlements, and the founding of new settlements.

Our indicators, windows and comparison set are constructions of this paper and are presented as such. The archive records what its observers saw; the choice of what to measure from those records is ours.

2. Sources and Method

The Contact Section Archive holds two relevant series. The Sublimation observation series lists every civilisation that Contact's observers recorded as having left the Real, with their registers, population estimates and settlement lists (Contact Section Archive, 1893 CE). The continuous observation register covers Involved civilisations kept under observation without a gap of more than ten years (Contact Section Archive, 1895 CE). Between them the two series list 31 civilisations observed to a standard we could use. Of these, 9 were recorded as having departed and 22 were not. One departed civilisation was excluded because observation lapsed for 140 years inside its final 400. Ten of the 22 that had not departed were excluded because no unbroken 400-year stretch of observation existed. The analysis therefore covers 8 departing and 12 comparison civilisations.

A departure is a civilisation the archive classifies as having left the Real. The archive's observers apply that classification where no remains of a failed population were found, no migration was recorded and the civilisation's characteristic signals ended abruptly (Ostrelle and Hollisent, 1841 CE). We adopted the classification as recorded and did not review it. The departure year is the year in which observers recorded that no resident population remained, given to the nearest five years. The span from the first recorded mass cessation of residence to that year ranged from 3 to 12 years across the eight cases (median 6.5), and Table 1 gives each case, labelled A to H in order of departure. Observation in all eight cases was passive. The records show no Contact action intended to hasten or delay any departure, though restricted holdings, which we did not consult, may contain material the open series do not.

For each civilisation we took a window of 400 years ending at the departure year, or for comparison civilisations ending at a year drawn by lot between the earliest and latest departure years, subject to the window lying inside continuous observation. The comparison windows end between 2210 BCE and 1290 CE. Each window was cut into eight 50-year bins. The split between the first four bins and the last four was fixed at the midpoint before any bin was examined, and no break point was searched for.

Three indicators were taken from the registers. Population growth is the mean annual percentage change in the estimated population over each bin. Population estimates come from remote survey and carry a stated margin of about 15% at each date (Verthane, 1783 CE). Series SO-2 is a compilation that extends back through the records of earlier observers, so the paper rests on no stated age for Contact, and the five departures dated BCE are taken as the series gives them. A common observance is a rite in which the register records that a majority of a settlement's adult residents took part at least once a year. Founding is the number of new settlements the registers list in a bin. In each bin we coded up to 30 settlement records for observance (all of them where the register held fewer, to a minimum of 16 per bin in every case analysed), giving 3,736 coded records across the 160 civilisation-bins. The first two authors coded independently. On a random sample of 160 records the two coders agreed on 144 (90.0%), with Cohen's κ = 0.71 (95% CI 0.58–0.85). The coding followed the register-reading procedure set out in Sorrelhaven (1860 CE).

The civilisation is the unit of analysis, so settlement records within a civilisation are not treated as independent. For each civilisation we computed the change in each indicator between the last four bins and the first four. For founding, the change is the log of (settlements in the last four bins + 0.5) over (settlements in the first four + 0.5). We compared departing and comparison civilisations with Welch two-sample tests on these changes and report 95% confidence intervals for the difference in means. An exact permutation test over all 125,970 ways of assigning 8 of the 20 civilisations to the departing group served as a check. With three indicators, we report Holm-adjusted p-values alongside the unadjusted ones. All statistics were computed in code from the coded registers.

3. Population Growth Before the Leaving

Departing civilisations grew more slowly in the last 200 years of the window, and comparison civilisations did not. Across the eight departing civilisations, mean annual growth fell from 0.44% in the first four bins to 0.16% in the last four, a change of −0.28 percentage points per year. Across the twelve comparison civilisations it moved from 0.43% to 0.45%, a change of +0.02. The difference in the change between the groups was −0.31 percentage points per year (95% CI −0.37 to −0.25; Welch t(17.7) = −11.06, p < .001; Holm-adjusted p < .001). The permutation test agreed (p < .001).

Individually, the civilisations separate cleanly. All eight departing civilisations showed a decline, ranging from −0.37 to −0.22 percentage points per year. The comparison civilisations ranged from −0.11 to +0.12, and 4 of the 12 declined. No comparison civilisation fell as far as the least affected departing one.

The bin means show where the decline lies (Table 2). In the departing group, growth held near 0.50% per year for the first two bins, then fell from the third bin (300 to 251 years before departure) to a mean of 0.09% in the last 50 years, with one small reversal between the sixth and seventh bins. The comparison group stayed between 0.41% and 0.51% in every bin. We tested no onset date, and the bin pattern is a description of means, not an estimate of when a decline began in any one civilisation.

Registers record population totals and not their components, so they cannot say whether births fell, deaths rose or residents moved to other settlements outside the observed region. Vandrasque and Lissome (1877 CE) read the slowing as lower fertility.

4. A Common Observance in the Settlements

A common observance became more frequent in the settlements of departing civilisations. Pooling the coded records, 167 of 752 (22.2%) in the first four bins reported one, against 242 of 742 (32.6%) in the last four. In the comparison civilisations the shares were 254 of 1,154 (22.0%) and 218 of 1,088 (20.0%). The test used the change within each civilisation, which averaged +10.3 percentage points among the departing and −1.5 among the comparison civilisations.

The difference in the change was +11.8 percentage points (95% CI +4.0 to +19.5; Welch t(16.9) = 3.19, p = .005; Holm-adjusted p = .011). The permutation test gave p = .013. All eight departing civilisations showed an increase, from +2.9 to +19.2 points. Of the comparison civilisations, 4 of 12 showed an increase, and their changes ranged from −19.9 to +19.9 points.

By bin (Table 2), the share in departing civilisations moved from between 19% and 25% in the first five bins to 32%, 39% and 38% in the last three. The comparison group varied between 15% and 26% without a trend. The coders' agreement was good but not perfect (κ = 0.71), and disagreements were not evenly spread: the coders disagreed on 16 of the 160 records, with the first coder recording an observance where the second did not in 9 and the reverse in 7.

The records give the form of the observances only as the registers describe them, and we did not code their content. Drevoly (1856 CE) proposed that shared observance binds a society as it approaches a decisive change. Our result shows that observance of this kind spread in the settlements of departing civilisations. It does not show what the observances meant to the people who took part.

5. Settlement Founding and the Departure Itself

Settlement founding did not separate the groups. Departing civilisations founded 170 settlements in the first four bins and 152 in the last four, a drop of 11%. Comparison civilisations founded 271 and 304, a rise of 12%. Per civilisation, this is about 10.6 and 9.5 foundings per century for the departing group, and 11.3 and 12.7 for the comparison group. The mean log ratio was −0.10 among the departing and +0.12 among the comparison civilisations. The difference was −0.21 (95% CI −0.48 to +0.05), which corresponds to a ratio of ratios of 0.81 (95% CI 0.62 to 1.05; Welch t(14.9) = −1.73, p = .10; Holm-adjusted p = .10). The interval includes a ratio of 1. Six of the eight departing civilisations and five of the twelve comparison civilisations founded fewer settlements in the last four bins than in the first four.

Table 3 sets the three results side by side. The direction of the founding result matches the other two, a lower rate of outward growth among those that departed, but the data cannot distinguish it from no difference, and we report it as unresolved.

The departures themselves were brief by the standard of the windows. Observers recorded 3 to 12 years between the first mass cessation of residence and the last resident (Table 1). In case H, the third author was in attendance for the final two years and kept its own logs, so observation of that case was otherwise remote only in its earlier years. We used them only to check the archive's population estimate for the last bin of that case, and the two agreed within the archive's stated 15% margin.

None of the eight registers records an attempt by the departing civilisation to communicate a reason to observers, and none records contact with them in the closing years. We note this as a fact about the registers. The registers say nothing about whether such contact was attempted and unrecorded.

6. Discussion

Two of the three indicators distinguished the eight departing civilisations from the twelve comparison civilisations in their final 200 years. Population growth fell in every departing civilisation, by about 0.28 percentage points per year on average, and in none of the comparison group by as much. A common observance became more frequent in the settlements of every departing civilisation, though the comparison group overlapped and the interval is wide. Settlement founding did not separate the groups. These are findings about the records of eight civilisations.

The comparison group matters for how the earlier proposals should be read. A slowing of growth and a spread of shared observance were each proposed from departing civilisations alone. The slowing of growth appears in all eight departing civilisations and in none of the twelve comparison civilisations to the same degree. The rise in observance appears in all eight departing civilisations and in 4 of the 12 comparison civilisations, so it is a weaker discriminator. The comparison civilisations were observed over windows of the same length. Twelve is a small group on which to rest any conclusion about what is ordinary for long-lived societies.

What the records cannot show is why these changes occurred or whether they played any part in departure. The patterns may be consequences of a decision to depart, accompaniments of it or independent of it. They may also be features of civilisations that happen to be the kind that depart. Nor can we tell whether the indicators would have identified departure in advance. They were identified by looking back from known departures, and we did not test them on civilisations whose futures were unknown.

One caution applies to the comparison group. A civilisation not recorded as departed by 1895 CE may yet depart, and the comparison windows may therefore contain civilisations early in the same process. If the indicators do track an approach to departure, such cases would tend to narrow the differences we found. We have no record of them to correct for.

7. Limitations of the Record

The classification of a civilisation as departed is the archive's, and it rests on the absence of remains, migration and signals. A reviewer could reasonably ask whether some of the eight were extinctions or migrations that observers mistook for Sublimation. We cannot exclude that, and the small number of cases means one misclassification would move the results.

Observation was otherwise passive and remote. Population estimates carry a margin of about 15% at each date. We treat that margin as a systematic bias that is constant within a civilisation and so cancels in the within-civilisation changes we analyse, with only a smaller random component remaining. Growth rates derived from them over 50-year bins are less certain than the two-decimal figures in the tables suggest. We report two decimals for consistency across tables and not as a claim of that precision. The registers of the earliest departures were reconstructed from the fewest surviving observer notes, so their population estimates are the least secure.

The coding of observance depends on the registers' own wording, which differs between observers and eras. Agreement between the two coders was good, but a single coding scheme applied to registers written by many observers may carry observer differences into the shares. Settlement records within a civilisation are not independent, which we handled by analysing civilisation-level changes at the cost of power.

Comparison windows were placed by lot within the span of departure years, so the groups are matched in era only on average. We did not consult the restricted holdings of the Special Circumstances Case Archive. Contact and Special Circumstances both keep some material out of the open series, and a record of action held there would alter our statement that Contact did not intervene. Survivorship also shapes the sample: 10 of the 22 non-departing civilisations and 1 of the 9 departing ones were dropped for broken observation, and civilisations that were harder to observe may differ in ways we cannot see. The records end at departure by design, and nothing in them bears on what followed.