The Longer the Cover, the Longer the Return: Predictors of Readjustment Time in 171 Post-Deployment Records of Contact Field Agents, 1702–1894 CE

1. Introduction

Contact meets other civilisations, watches their development and, where it judges it right, steers them toward peaceful paths. Much of that work is done by people who go and live among them. Contact field agents are volunteers, they go by choice, and they may come home when they wish. A deployment among a society that has not left its home world can last a decade or more, and some agents spend it as members of the local population, with their origin unknown to their neighbours (Brelmott, 1801 CE).

Coming home is not always quick. An agent returns to a society in which nobody owes anyone a living and nobody holds rank, after years among people for whom none of that is true. Contact's clinicians have long said that some agents take a season to settle and others take years. The published literature describes this in individual cases (Quillamar, 1824 CE) and in essays on the loss that follows a life lived under another name (Neldrasse, 1871 CE). It does not say what predicts the difference across a body of agents, and so does not tell Contact whom to watch most closely, or what to change.

This paper asks that question of a record series. The vantage of the analysis is 1900 CE. Every deployment in the series ended before 1895 CE, and the series was closed for this study in 1899 CE. We ask how long agents took to readjust after extended deployments, and which features of the deployment were associated with a longer or shorter return: its length, how deeply the agent was embedded in local life, whether anyone from the Culture went with them, whether they had been deployed before, and their age at departure.

2. Methods

The records come from the returns files of the Contact Section Archive (Contact Section Archive, 1702–1894 CE). When an agent returns from a deployment of two years or more, a monitoring clinician offers follow-up, and the file records the deployment, the agent's account of it and the clinician's dated notes. Follow-up is by invitation. A clinician writes down what the agent chooses to say and does not read the agent's mind, in keeping with the long-standing practice of monitoring by consent (Tarvenholt, 1763 CE). The series that we call the PD register is our own assembly of these files for this study, and the selection rules below are ours.

Within it, 203 deployments ended between 1702 and 1894 CE. Twenty-four carry a note that the file is held in a separate restricted archive. We saw only the finding aid for these (Contact Section Archive, 1895 CE), and we did not read them or draw on them in any way. A further 8 were still open when the series was closed for this study in 1899 CE, that is, monitoring had not yet been ended by agreement. We left them out because their outcome is unknown. That leaves 171 deployments by 121 agents: 82 agents contribute one record, 28 contribute two and 11 contribute three.

Our outcome is readjustment time, the number of whole months from the agent's return to the date on which the clinician and agent agreed that monitoring could end. The clinician's notes give the agent's own account of when ordinary life had resumed: chosen occupations taken up again, company kept, and the deployment no longer the main subject of conversation. The definition rests on agreement between two people and is not a clinical threshold. Closure dates are given in whole months in the files.

Five features were coded from the files. Deployment length is in years. Depth of cover has three levels. Observers lived apart from the local population, with contact limited to what the work required. Participants took part in local life openly, as known visitors. Full-cover agents lived as members of the local population and concealed their origin. Companionship records whether another Culture citizen, whether pan-human or drone, was present for most of the deployment. Prior deployment records whether the agent had returned from an earlier extended deployment in the analysed series. Age at departure is in years.

Two of us, Veltrecourt and Doanthe, coded depth of cover independently for a random subsample of 60 records, then settled all disagreements by discussion before coding the rest. They agreed on 53 of the 60 (88%), a Cohen's kappa of 0.82. Companionship, prior deployment, length and age are recorded as such in the files and were transcribed without interpretation.

Because 89 of the 171 deployments belong to agents with more than one deployment in the series, records from one agent are not independent. We modelled the natural logarithm of readjustment time by ordinary least squares with all five features entered together and standard errors clustered by agent (121 clusters). Intervals use a t distribution with 120 degrees of freedom. Exponentiated coefficients are ratios of geometric mean readjustment time. Length is scaled per five years, centred on 11 years, and age per 50 years, centred on 150. The distribution of readjustment time is right-skewed, so we describe it with medians and interquartile ranges. Because the register is archival, we report estimates to two decimals and do not read more precision into them than the files carry.

The third author, Distaffe, is a Contact clinician who wrote the monitoring notes for 11 of the 171 records, all from the final 15 years of the series. She did not code any variable. The other two authors, who work in the Archive, compiled the register and did the coding.

3. Results

Deployments ended between 1702 and 1894 CE. The median length was 11 years (interquartile range 8–16.5, range 2–40) and the median age at departure was 185 years (126–226, range 60–288). Sixty-six deployments (38.6%) were by observers, 61 (35.7%) by participants and 44 (25.7%) by full-cover agents. Seventy-nine (46.2%) had a companion present. Fifty were second or third deployments (29.2%), and the other 121 were an agent's first.

Readjustment took a median of 8 months (interquartile range 5–12, range 2–43). It reached a year or more in 44 records (25.7%) and two years or more in 9 (5.3%). Before adjustment, the median was 6 months for observers (4–10), 8 for participants (5–11) and 10 for full-cover agents (7–17). It was 6 months for agents with a companion (4–9) and 9 for those without (5–13). A companion was present in 28 of the 66 observer deployments (42%), 28 of the 61 participant deployments (46%) and 23 of the 44 full-cover deployments (52%).

The adjusted model is in Table 1. Full cover was the strongest association: full-cover agents took 80% longer than observers (ratio 1.80, 95% CI 1.43–2.27, p < .001), and participants 31% longer (1.31, 1.07–1.61, p = .01). Each additional five years of deployment was associated with 9% longer readjustment (1.09, 1.03–1.16, p = .003). Having a companion was associated with 32% less time (0.68, 0.57–0.82, p < .001). Prior deployment (0.96, 0.80–1.16, p = .69) and age at departure (1.02 per 50 years, 0.94–1.11, p = .61) showed no clear association, and the intervals for both include a modest effect in either direction. The model explained 25% of the variance in log readjustment time, and the residual standard deviation was 0.60 on the log scale.

For length, depth of cover and companionship, the direction of the raw differences in medians was the same after adjustment.

4. Discussion

In these records, the features of a deployment recorded in the file, how long an agent is away and how completely they live as someone else, predict how long the return takes. An agent who spent a decade as a member of another people took, on average, nearly twice as long to settle as an observer of the same age on the same length of deployment. The association with length is smaller. Each five years away added under a tenth to the time.

Agents with a companion returned sooner. The ratio of 0.68 has an interval that excludes 1. We cannot say that the companion caused the shorter return. The files do not record who decided that an agent should be accompanied, and Contact may have sent companions more readily with agents it expected to find the work harder, which would hide a benefit rather than create one. Equally, an agent who is able to have a companion may differ from one who is not in ways that the files do not record. The earlier literature describes companions as a channel through which an agent stays in touch with Culture manners during a long absence (Sennovich, 1858 CE); the present records agree with that description in direction but do not test the mechanism.

The absence of a clear association with prior deployment is easy to over-read. The interval for this estimate runs from 0.80 to 1.16. The records cannot show that experience does not help, only that any help is small relative to the other features. The same holds for age at departure, within the range of 60 to 288 years covered by the register.

For Contact's clinicians, the practical reading is about attention. Directing attention first to full-cover and long deployments follows the associations found here. Whether additional support for those agents shortens the return is a question these records cannot answer, because every agent here received the follow-up offered at the time.

5. Limitations

The records are observational, and the features were not assigned. Prior deployment is counted only from the analysed series, so earlier deployments whose files are restricted are not counted. Length may also have been shortened or extended by the agent's own state, so that an agent who was struggling, or settled, may have ended a deployment sooner or later; the association with length could therefore run in either direction. Companionship and depth of cover were chosen for reasons the files do not record, and agents and clinicians may have had views about readjustment that influenced both. We did not model these choices.

Agreement between an agent and a clinician about when ordinary life had resumed defines the outcome. Agents who chose to end monitoring early are counted as having readjusted quickly, and agents who stayed in contact with clinicians longer than they needed are counted as slow. We cannot separate these from the experience of readjustment itself, and the definition may have shifted over the 192 years of the series.

Eight records were still open at the close of the series. Their agents had not been monitored to closure, so they are probably the slowest returns, and leaving them out will have understated the length of the longest returns and may have biased some of the associations. The 24 deployments held under restricted access were not seen; they may differ systematically from those we analysed, and we say nothing about them.

Finally, the records cover only agents who returned and accepted follow-up. They give no information about agents who did not come back or did not ask for monitoring, and the depth-of-cover variable was coded from notes that vary in detail from one clinician to another, with a kappa of 0.82 on the subsample that was double coded. The model describes association in this register and should not be applied to deployments of a different kind.