What a Mind Calls Itself: Register, Length and Sentence Form in 361 Self-Chosen Names, by Ship Class and Hub, from the Chiark Orbital Hub Name Register, 1100–1740 CE

1. Introduction

A Mind chooses its own name. No one assigns it and no authority approves it, and Contact reads a name as the first thing a Mind says about itself. The names are famous for what they do with this freedom: they run to long sentences, dry remarks, double meanings (Isseldane, 1704 CE) and pointed courtesies, and visitors from other civilisations often take them for jokes. Contact has long regarded them as a form of self-description in which the Mind decides how it wishes to be addressed and how it wishes to be understood.

These qualities have been described far more often than they have been counted. Accounts of Mind naming rest on collections of striking examples, and striking examples are by definition the unrepresentative ones. Vennerholt's survey of 1604 CE is a catalogue of striking cases and does not give class-by-class frequencies (Vennerholt, 1604 CE). We do not know whether sardonic names are common or merely memorable, whether the long sentence-names that dominate anecdote are typical of any class, or whether the Minds that run Orbitals name themselves differently from the Minds that travel.

This paper offers a first count. We use a register of self-chosen names kept by one Hub and ask three questions. Do names differ in register, that is, in the attitude the name takes towards its hearer, between classes of Mind? Do they differ in length? And do they differ in the grammatical form of the name, as a phrase or as a clause? The questions are put to five classes: General Systems Vehicles (GSV), General Contact Units (GCU), Limited Offensive Units (LOU), General Offensive Units (GOU) and Hubs. We write in 1750 CE, and the register we use closes with the volume of 1740 CE.

Class matters to the question because the classes differ in what their Minds do and in how they meet others. A GCU is built to meet strangers; a Hub is the Mind of an Orbital and is addressed daily by its residents; a dedicated warship is built for a different purpose again, and follows the same custom of self-naming as any other Mind (Pomeroyne, 1738 CE). Whether these circumstances shape the names is an open question, and we treat it as one.

2. Methods

Our source is the register of Mind self-names kept by Chiark Orbital Hub, which we cite by its holding series (Chiark Orbital Hub Records, 1100–1740 CE). A Mind based at Chiark or visiting it may enter its name in the register by choice, with the date and the class it gives for itself; many do, and many do not. The register is a record of what Minds elected to enter and makes no claim to list every Mind that has passed through (Calquarrin, 1688 CE). Hubs of other Orbitals entered their names by message to the Chiark Orbital Hub Archive. Each entry holds the name as the Mind gave it, rendered in the register's working language, together with any note the Mind chose to add.

The register holds 392 entries for the window. We excluded nine entries from Rapid Offensive Units, too few for a class comparison, seven repeat entries in which a Mind had entered the same name again (we kept the first), and fifteen entries whose recorded name was illegible or lost to damage in the volume. That left 361 names, each from a different entry: 88 GSV, 121 GCU, 46 LOU, 32 GOU and 74 Hub. The 74 Hub names come from 74 different Orbitals, Chiark's own among them. No entry from a Rapid, Limited or General Offensive Unit predates 1328 CE; the offensive-unit classes are therefore compared over the years from 1328 to 1740 CE, whereas the other classes span the whole register. For the contrast between ships and Hubs, we group the four ship classes as 287 ships. For one exploratory contrast we also group LOU and GOU as 78 offensive units. The class prefix (GSV, GCU, LOU, GOU) and the suffix Hub belong to the Mind's designation and are excluded from every count of words. For a Hub, the stem is whatever designation the Hub entered in the register, and we make no claim about how Hub designations arise generally (Ostrevane, 1732 CE).

Every name was coded on three variables. Length is the number of words in the recorded rendering, counted by the conventions of Dunmarrow (1721 CE). Sentence form has four values: noun phrase, finite clause, imperative or question, and fragment or other. Register has five mutually exclusive values, assigned by the dominant attitude of the name towards its hearer or its own situation: sardonic, a wry or dry remark; layered, a literal reading and an allusive one that must both be held; courteous, a name that welcomes, thanks or honours; plain, a descriptive or declarative name with no comment; and other. We illustrate the scheme with names of our own devising, none drawn from the register: a name such as Nobody Asked The Cargo would be coded sardonic, The Long Way To A Short Answer layered, Welcome Is The Whole Ship courteous and Quiet Water plain.

One of us coded all 361 names from the name itself, with the Mind's own note available but not used. A second coded a random 120 of them (one third) independently and blind to the first, and disagreements on those names were settled by discussion. We report Cohen's kappa, which for several categories and two coders is the measure Esslin (1699 CE) recommends. Agreement was 97 of 120 for register (kappa 0.75) and 112 of 120 for sentence form (kappa 0.90). Register is the harder variable, and its categories blur at the margin between sardonic and layered.

The tests follow the variables. Register and sentence form are categorical, and several class cells are small, so we assessed class differences by a chi-square statistic referred to its permutation distribution (10,000 permutations of class labels), and report Cramér's V as effect size. Length is a small count, so we compared classes by the Kruskal–Wallis test and report epsilon-squared. Proportions carry Wilson 95% confidence intervals, and differences between proportions carry Newcombe intervals. After inspecting the full table we examined three contrasts, each in a form given by the result and not by an earlier hypothesis: courteous names in Hubs against ships, noun-phrase names in Hubs against ships, and sardonic names in offensive units against the other ships (GSV and GCU). We refer these to Fisher exact tests with a threshold of .017, or .05 divided by three. Because the contrasts were selected from the data, their p values are conditional on that selection.

3. Results

Names were shortest in Hubs and longest in General Systems Vehicles (Table 1). The overall median was three words (mean 3.24), and the pooled ships had a median of three (IQR 2–5, mean 3.67). Hub stems had a median of one word, and 64 of 74 (86.5%) were one or two words long, against 78 of 287 ships (27.2%). Names of seven or more words were confined to two classes: 11 of 88 GSV names (12.5%) and 4 of 121 GCU names (3.3%). Length differed across the five classes (Kruskal–Wallis H = 110.17, df = 4, p < .001, epsilon-squared 0.31). Among the four ship classes alone, length still differed, though less (H = 17.91, df = 3, p < .001, epsilon-squared 0.06).

Register differed across the five classes (Table 2: chi-square = 40.27, df = 16, permutation p < .001, Cramér's V = 0.17). Sardonic and layered names were the two most frequent registers overall, 101 and 98 of 361 names (28.0% and 27.1%), and together they made up 199 names (55.1%). Courteous names followed at 77 (21.3%), plain names at 58 (16.1%) and the residual category at 27 (7.5%).

The difference lay chiefly with Hubs. Courteous names were 26 of 74 Hub names (35.1%, 95% CI 25.2–46.5%) and 51 of 287 ship names (17.8%, 13.8–22.6%), a difference of 17.4 points (Newcombe 95% CI 6.3–29.4; Fisher p = .002). Among the four ship classes alone, register did not differ detectably (chi-square = 16.71, df = 12, permutation p = .16, Cramér's V = 0.14). Ship names were closer to one another than to Hub names: the Hub row alone contributes 19.18 of the 40.27 chi-square units (48%), and with the Hubs removed the statistic falls to the 16.71 just given, on 12 degrees of freedom.

Sentence form followed the same division. Over all 361 names, 120 (33.2%) were noun phrases, 142 (39.3%) finite clauses, 67 (18.6%) imperatives or questions and 32 (8.9%) fragments or other forms. The five classes differed (chi-square = 37.52, df = 12, permutation p < .001, Cramér's V = 0.19). Noun phrases were 42 of 74 Hub names (56.8%, 45.4–67.4%) and 78 of 287 ship names (27.2%, 22.4–32.6%), a difference of 29.6 points (Newcombe 17.0–41.3; Fisher p < .001). Among ships, sentence form did not differ detectably by class (chi-square = 11.46, df = 9, permutation p = .24).

One contrast concerns the offensive units. Sardonic names were 34 of 78 offensive-unit names (43.6%, 95% CI 33.1–54.6%) and 58 of 209 GSV and GCU names (27.8%, 22.1–34.2%), a difference of 15.8 points (Newcombe 3.6–28.2; Fisher p = .015). All three contrasts were post hoc. The two Hub contrasts clear the adjusted threshold comfortably; the sardonic one falls below .017 only narrowly, and its interval is wide. The omnibus test among ships found no register difference, so we treat this contrast as exploratory. The two offensive classes pointed the same way: 22 of 46 LOU names (47.8%) and 12 of 32 GOU names (37.5%) were sardonic.

4. Discussion

The clearest finding is a division between Hubs and ships. A Hub tends to give itself a short name, often a single word, in the form of a noun phrase, and more often than a ship it chooses a courteous one. The register cannot say why. A Hub is a Mind that residents address daily and from whom they ask things, and a short, welcoming name suits that standing; but nothing in our data tests that reading, and a Hub that runs an Orbital may simply have chosen differently for reasons the register does not hold (Ostrevane, 1732 CE).

Among ships, the names are more alike than the anecdotes imply. We found no detectable difference in register or sentence form across GSV, GCU, LOU and GOU, and the confidence intervals for individual cells are wide enough that moderate differences could be hidden. What did differ was length: General Systems Vehicle names were the longest, with a median of four words, and the only names of seven words or more came from GSVs and GCUs. The effect is small (epsilon-squared 0.06) and we offer no account of it.

The sardonic lean of the offensive units is the most tempting result and the least secure. Close to half of LOU names were coded sardonic, and the GOU figure was similar in direction. The contrast passed the adjusted threshold by a narrow margin, was selected after the table was seen, covers a shorter span of years than the other ship classes, and sits beside an omnibus result that found nothing among ships. We report it as a lead for a study designed to test it. We make no claim about the temperament of offensive-unit Minds, and the register gives no basis for one.

Sardonic and layered names together made up a little over half of the corpus. This agrees with the reputation the names have, and it adds a number to it. It also shows that the reputation is not complete: about one name in five was courteous, one in six plain, and a Mind that chooses a plain name has chosen it as surely as any other. A parallel study of drone aura displays asks the same question of another kind of citizen, whether a display is a chosen register or an involuntary readout, and finds most displays deliberate.

5. Limitations

The corpus is a voluntary register at a single Orbital. A Mind decides whether to enter its name, and Minds that favour entering may differ in temperament from those that do not, so the proportions describe registrants and not Minds in general. Hubs may be over-represented, because a Hub is addressed daily and can enter by message at any time, whereas a ship must pass through or contact Chiark. We cannot estimate how many Minds are missing, and we make no claim that the Culture's Minds as a whole are distributed as the registrants are.

Coding rests on the rendering in the register's working language, and not on the original forms. Word counts and register depend on how a name was rendered, and a layered name may lose a layer in rendering. Length in particular is the length of the rendering. Coders could see any note a Mind added but did not use it.

Names from Minds that entered on the same visit or in the same year may resemble each other, because Minds who meet often borrow phrasing and play on each other's choices. We treated names as independent. Our permutation tests inherit that assumption, and the true uncertainty is probably somewhat larger than the intervals suggest.

Offensive-unit names cover a shorter span, 1328 to 1740 CE, than the other classes, which span 1100 to 1740 CE, so class and period are partly confounded. We did not model period, and any difference in the offensive classes may reflect their era as much as their class.

Register is a human judgement. Agreement of kappa 0.75 is adequate and is not high, and the border between sardonic and layered is the least stable in the scheme. The coders are two readers from one research group, both at home in Chiark's idiom, and Mind names from other regions of the Culture may read differently to them. All class assignments are those the Minds gave for themselves, and we did not check them against any other record. The three contrasts were chosen after seeing the table, and their p values are conditional on that choice; dividing the threshold by three understates the number of contrasts that could have been chosen.