Muster and Rite: Recruitment, Garrison Distribution and Devotional Cohesion in the Fish Speaker Host, 13630–13690 AG

1. Introduction

The muster was read in the devotional hall before the first meal. A registrar stood at a lectern with the garrison ledger and called roll numbers, not names, and each woman answered with a short formula of devotion to the God Emperor. Of the 412 numbers called that morning, 409 were answered; the registrar set the three absences against the sick list and turned the page. We attended that muster at the largest garrison on Arrakis in 13684 AG. It holds the paper's three subjects in one room: a ledger that counts the host, a hall that gathers it, and an answer addressed to a person most of those giving it would never meet.

We write c. 13740 AG, in the years after the God Emperor's death. The fieldwork reported here was carried out in 13684–13690 AG under an Imperial licence that barred publication while he lived, and the notebooks and recordings reached the Imperial Archives at Kaitain only afterward (Academy Field Team on the Speakers' Host, 13684–13690 AG). We have re-read and re-coded them; the muster and intake statistics are our own compilation from the same archives.

The Fish Speakers were the God Emperor's army of women. His stated reason was that women are less given to the male aggression he meant to tame, and the court record preserves it (Imperial Recorder, 10300–13690 AG). The host was also a religious body, bound to him by devotion as much as service. Its size under the Peace he imposed has been estimated from ration returns alone (Vorlaine, 13706 AG), and its internal order has received little study.

What kept such a host together? The answer could lie in whom it admitted, in the rite that formed members after entry, or in an authority that gave every order a source beyond the officer who spoke it. The Sardaukar supply a comparison: earlier work on their training found that the harshness of Salusa Secundus both culled the weak and improved the survivors, so that a place did work no instructor could do. We ask whether any such engine operated for the Fish Speakers.

2. Field Setting and Method

At the muster of 13688 AG the host numbered 6,840 members in 31 garrisons. We sort the garrisons into Arrakis, where the God Emperor kept his seat, a near ring within twenty days of scheduled Guild passage from Arrakis and a far ring beyond. Every garrison has a devotional hall, where the host performs its daily rite, and a hall-keeper, an officer separate from the commander, who keeps it.

Annual muster returns survive for 49 of the 61 years 13630–13690 AG (Registrar of the Speakers' Host, 13630–13690 AG). The registrar's intake and novitiate registers cover the 30 intake years 13650–13679 AG and follow 2,184 entrants to 13690 AG (Registrar of the Speakers' Host, 13650–13690 AG). As a check we used the method of earlier work by one of us (Draybourne, 13712 AG): ration-derived and muster strengths agreed within 6% in 27 of the 31 garrisons and within 11% in the rest.

The field record comprises nine visits to six garrisons, two on Arrakis and two in each ring, totalling 31 weeks: about 400 hours of observation of musters, drill and the daily rite, the hall-keepers' attendance books, and 63 recorded interviews with 41 serving members, 14 retired members and eight hall-keepers. In 38 interviews no officer was within earshot.

Transcripts were coded for reasons given for entering, reasons for staying and the source of orders. Garrison strength, a count, is modelled by negative binomial regression; ratification is compared across routes by a χ² test and odds ratios; departures are rates per hundred member-years, compared as rate ratios with intervals on the log scale. Member-years from one garrison are not independent, so we widen those standard errors by 1.3, the inflation implied by between-garrison variation.

3. Findings: Entry and the Novitiate

The intake registers record three routes into the host. In household dedication a family vowed a daughter to it. In application a woman presented herself at a devotional hall. In transfer a woman serving in the God Emperor's household was released to the host (Orsolane, 13688 AG). No entrant was recorded below sixteen. The 30 cohorts held 2,184 entrants: 1,126 by dedication, 842 by application and 216 by transfer.

Every entrant served a twelve-month novitiate closed by a ratification rite, which the register treats as the mark of full membership. Overall 1,745 of 2,184 (79.9%) were ratified, but the share differed by route (χ²(2) = 35.5, p < .001). Applicants had roughly twice the odds of ratification of dedicated entrants, a gap of 10.9 percentage points (95% CI 7.4–14.3). Transfers fell between the two, and their odds ratio against dedication does not exclude no difference. The registers give no reasons for the 439 who were not ratified.

Interviews add a convergence the registers cannot show. Asked why they stayed, 49 of the 55 serving and retired members (89%) named the God Emperor's presence in the rite, or the hall itself: 21 of 24 dedicated entrants, 20 of 22 applicants and eight of nine transfers. Several who entered under a family's vow said the novitiate's daily instruction had made the vow their own.

4. Findings: Garrisons and the Seat

The return of 13688 AG counts 6,840 members in 31 garrisons. Earlier estimates, made without the muster returns, ran from about 5,000 to 9,000 (Vorlaine, 13706 AG); the muster count lies inside that range. Arrakis held nine garrisons and 44.7% of the host, the near ring 38.2% and the far ring 17.1%, and garrison size fell with distance, as the table shows.

In a negative binomial regression of garrison strength, each additional devotional hall in the district, counting the garrison's own, multiplied expected strength by 1.32 (95% CI 1.12–1.55, p < .001), with passage time and civil population held fixed; districts held one to five halls. Each further ten days of Guild passage multiplied strength by 0.81 (95% CI 0.68–0.97, p = .02). Civil population showed no detectable association: a tenfold larger population multiplied strength by 1.12 (95% CI 0.74–1.69, p = .60). Disturbances noted in the returns added nothing (rate ratio 1.03 per disturbance, 95% CI 0.95–1.12).

The host was therefore stationed around the seat and around places of devotion, and not around people or trouble.

5. Findings: Authority, Rank and the Daily Rite

Each garrison ran on two lines of officers. The commander directed duty, drill and supply; the hall-keeper led the dawn rite and kept the attendance book. Attendance at the dawn rite averaged 93% of off-duty members across the sampled weeks. The books record 52 lapses classed as unexcused; the hall-keeper alone dealt with 47, and five were referred to the commander.

Senior rank followed the hall. The surviving returns record 118 appointments to the two highest of the host's five grades. Of these, 79 (66.9%, 95% CI 58.5–75.4) went to women who had served as hall-keepers, a group that made up about 14% of ratified members.

Appointments were also tied to the apex. A separate notation in the returns records that the court of the God Emperor confirmed the appointment. It appears against 103 of the 118 senior appointments (87.3%), and all 15 without it date from before 13646 AG. Interviewees likewise place final authority in a person. Asked who could countermand a garrison commander, 47 of 63 (74.6%) named the God Emperor himself, 11 (17.5%) the hall-keeper and five (7.9%) a superior officer.

Cohesion shows in who left. A departure for cause means desertion, or discharge for a lapse the register classes as loss of devotion. There were 257 among the 1,745 ratified members over 29,640 member-years. Members in garrisons with a hall in daily use left at 0.69 per hundred member-years (168 departures in 24,190); members on detached posts of over a year, with no hall within a day's travel, left at 1.63 (89 in 5,450). The rate ratio was 2.4 (95% CI 1.7–3.3).

Commanders might have sent less devout women to detached posts. The 396 members who served in both kinds of post address this, because each is compared with herself: 0.74 per hundred member-years on hall posts and 1.74 on detached posts, a rate ratio of 2.4 (95% CI 1.4–4.0). Remoteness alone did not matter. Among hall-posted member-years the rate was 0.68 on Arrakis and 0.72 in the far ring (rate ratio 1.06, 95% CI 0.61–1.86).

6. Discussion

The Sardaukar offer the instructive contrast. Earlier work found that the harshness of Salusa Secundus did part of the court's work, removing weaker children before recruitment and improving the survivors, which is why the court never moved its training elsewhere. The Fish Speakers show no counterpart. Remote posts did not retain members better: with a hall, far-ring members left at about the Arrakis rate, and without one, members left at more than twice it. The Sardaukar owed their quality to where they were raised; the Fish Speakers owed their cohesion to what they did each morning wherever they were posted.

Authority in the host also contrasts with the case of Muad'Dib, where a prepared legend passed to those who recognised it and then to a priesthood its subject did not fully command. Among the Fish Speakers the line from garrison to sovereign stays short: interviewees name the God Emperor as the last source of orders, and the court's confirmation is recorded against nearly nine in ten senior appointments. The hall line supplies most senior officers, yet the confirmation keeps each appointment in the sovereign's hand.

The God Emperor's stated reason for a host of women concerns the kind of soldier he wanted, and our evidence does not test it. It bears on what he built around the choice: a selective entry, a hall in every garrison, and a sovereign who remained the source of rank and orders. Selection explains why ratification differs by route. Convergent accounts of why women stay could result from selection alone, but the within-member contrast could not: the same women left at about 2.4 times the rate away from a hall. Members were admitted by selection and kept by practice.

7. Limitations

Muster returns survive for 49 of 61 years, and the missing years may not be missing at random. The intake registers record entrants, not the women turned away at the hall door, so the first filter is unmeasured. Departure for cause is the registrar's own classification, and the registrar's office had an interest in recording devotion. Detached posts were assigned by the host; the within-member contrast argues against assignment by devotion without excluding it. Twenty-five of the 63 interviews were held with an officer within earshot, and answers about the God Emperor given to outside scholars cannot be treated as unprompted. The regression rests on one year and 31 garrisons, and the series closes in 13690 AG, so the record cannot show the host's later course.

Within these limits the evidence supports four conclusions. The host admitted women by dedication, application and transfer, and ratified about four in five of them after a year. It stood around the seat and the devotional halls. Its authority ran through a command line and a hall line that met in the God Emperor, who confirmed most senior appointments. Its members stayed because of a daily rite: departures for cause were about 2.4 times as frequent where no hall was in reach, within the careers of the same women as well as across the host. The Fish Speakers were a selected body held together by practice, under an authority built to stay personal.