Catalogue Cross-Reference Integrity in the Jedi Archives: A Sample Audit of 1,420 Entries from the Released Historical Catalogue, 400–350 BBY

1. Introduction

The Jedi Archives on Coruscant are the oldest and most completely organised collection of records known to Republic scholarship, and the released historical catalogue is the window through which most galactic historians work with them. The catalogue is an abridged release: the keepers of the Archives, from an office within the Jedi Temple, publish summary entries for records whose full text remains in the Archives themselves and is accessible only on a request procedure that the released catalogue does not describe. The catalogue is organised by a running numeration that has, in principle, been stable for the whole of its history, and entries routinely cite one another by catalogue number. The cross-references are the practical fabric of the catalogue: they turn a list into a usable record.

No audit of the released catalogue's cross-reference integrity has been published, so far as we have been able to establish, in the current century. The question is a plain and necessary one for the historiography. If a historian cites "entry 384-C/2 cites entry 371-C/9", then on the strength of that one cite the citation is the record. If the cite fails, so does the claim. The integrity of such a chain in a working historical argument depends on the base rate of successful resolution; a 95% rate is a different world from an 85% rate for arguments of any length.

This paper reports an audit of a stratified random sample of 1,420 catalogue entries from the released historical catalogue for the fifty standard years from 400 to 350 BBY, chosen because the period is long enough to span a documented change in catalogue practice and short enough to audit by hand. We ask three questions. What proportion of outbound cross-references in a sampled entry resolve to the entry they name? What is the typology of the failures? And does the failure rate vary across subject fields and across the period of the sample? The vantage is 25 BBY.

2. Sources and Method

The sample frame is the released historical catalogue of the Jedi Archives for 400 to 350 BBY, deposited in the Jedi Archives (Jedi Archives, 400–350 BBY, released historical catalogue series). The catalogue contains, by our count on the deposited index, 41,228 issued entries in the fifty-year window. We stratified by five-year block and by seven subject fields used in the catalogue's own indexing: philosophical commentary, historiographic record, biographical, legal and ceremonial, early starfaring, planetary observation, and comparative religion. Within each stratum we drew a random sample proportional to the stratum size, with a floor of fifteen entries per stratum. The final sample contains 1,420 entries, which we hold is adequate for the research questions at a target standard error below one percentage point on the headline resolution rate.

Each sampled entry's outbound cross-references were extracted. A cross-reference is any reference by catalogue number to another entry in the same release. Reference by name alone to a non-catalogued source was coded as a non-catalogue reference and excluded from the resolution analysis. The 1,420 sampled entries yielded 3,864 outbound cross-references, a mean of 2.72 per entry. We did not treat multiple cross-references within a sampled entry as independent observations and we report cluster-robust confidence intervals, with the sampled entry as the cluster.

Resolution was tested by hand. A cross-reference was coded as resolved if an entry under the named catalogue number existed in the released catalogue and its summary content, as far as the released catalogue permits comparison, was concordant with the context of the citation. A cross-reference was coded as resolved-with-correction if an entry under a different number existed whose content was concordant and whose number differed from the cited one by rules we document in Section 3. A cross-reference was coded as absent if no entry under the named number existed and no near-number entry was concordant. A cross-reference was coded as ambiguous if two or more candidate entries were concordant with the citation and we could not choose on the released text alone. Three of us coded each cross-reference independently; agreement across all three was 93.1% and the remaining cases were decided by majority with a documented reasoned minority.

Subject field was taken from the citing entry's own classification in the released index, with the seven-field scheme described above. Period was taken from the citing entry's stated date, in the catalogue's own BBY numbering, binned into ten five-year blocks. Where a citing entry's date was itself cross-referential and could not be dated on the released text, the entry was retained in the period analysis under the latest attributable block, with a sensitivity analysis dropping those entries.

3. Resolution Rate and the Typology of Failures

Of 3,864 outbound cross-references, 3,224 (83.4%) resolved to the named entry, 356 (9.2%) resolved to an entry under a different number, 178 (4.6%) named an entry absent from the released catalogue, and 106 (2.8%) were ambiguous. The clustered 95% confidence interval on the headline resolution rate is 81.9% to 84.9%. We treat the four categories as a typology, and the three failure categories as separate problems with distinct implications for historiography.

Resolved-with-correction cases, at 9.2%, are the largest failure category and the most instructive. In 311 of the 356 cases, the corrected number differed from the cited one by a single digit, with the catalogue's own transposition patterns (a reversed pair, a shifted group separator or a misread terminal check-letter) accounting for 278 of the 311. In 32 of the 356 cases, the corrected number lay in an adjacent subject-field band whose numeration partly overlaps that of the cited field; those cases are concentrated in biographical entries whose subjects also appear in early starfaring. We treat the resolved-with-correction category as evidence of scribal, not archival, failure: the record exists, the entry exists and the referencing entry's content is right about the record it meant to cite.

Absent cases, at 4.6%, are the hardest to interpret. The released catalogue is itself an abridgement, and the keepers of the Archives have in several periods withdrawn entries from the released series without altering the numeration of the surrounding entries. An absent cross-reference may therefore name a withdrawn entry, a never-released entry or a genuinely missing one. We have no means on the released text alone to distinguish the three. In 41 of the 178 absent cases we found a reference in a later entry of the released catalogue to a withdrawn entry under the same number; in those cases the absent cross-reference almost certainly cites a withdrawn entry and not a missing one. In the remaining 137 cases we cannot say, and we treat them in the main table as absent without inferring withdrawal.

Ambiguous cases, at 2.8%, are concentrated in two patterns. The first is a catalogue number that was reissued at a cataloguing reform in the middle of the sample period (389–387 BBY), when a block of pre-reform entries were given post-reform numbers; a cross-reference written before the reform and read after it can point to either entry. The second is a cross-reference whose context is thin enough that two concordant candidates both fit; this is a problem of the citing entry rather than the catalogue, and it concentrates in philosophical commentary.

4. Variation Across Subject Fields and Across the Period

The headline resolution rate masks a mild but recognisable variation across the seven subject fields. The best-resolving fields are legal and ceremonial, at 90.1%, and comparative religion, at 88.4%, with narrow intervals around each. The weakest-resolving fields are early starfaring, at 76.3%, and biographical, at 77.1%; both failure categories are elevated in those two fields, and the resolved-with-correction category is heaviest in early starfaring. The two best-resolving fields are also those with the least activity in the sample period, and the two weakest are those with the most activity and the greatest overlap with the catalogue's reform.

Variation across the ten five-year blocks is monotone and modest. The resolution rate rises from 79.8% in the first block (400 to 396 BBY) to 86.4% in the last (354 to 350 BBY), with a monotone trend in nine of the ten block-to-block transitions. The improvement is accompanied by a fall in the resolved-with-correction category rather than in the absent or ambiguous categories, consistent with a tightening of the catalogue's internal scribal practice over the period rather than with any change in the keepers' release decisions. We note that the catalogue's own cataloguing reform of 389 to 387 BBY sits inside the improving trend and does not interrupt it.

The implication for a historian is quantitative. On the audit's own figures, a point claim resting on a single catalogue cross-reference is right about five times in six at the overall rate and better than that in the two strongest fields. A claim that chains two cross-references, read as independent at the audit's rates, is right about seven times in ten; a claim that chains three is right only three times in five. For arguments of greater length, the base rate of successful resolution is the dominant source of uncertainty, and no amount of care at any one link can repair it.

5. Discussion

The audit yields three statements for the historiography. The released historical catalogue of the Jedi Archives has a cross-reference resolution rate near 83% in the audited period, with meaningful variation across subject fields and a monotone improvement over time; the dominant failure mode is near-number scribal error, which the catalogue's own numeration patterns largely account for; and absent cross-references, which are 4.6% of the total, cannot be distinguished on the released text alone from entries that the keepers have withheld or never released. The third statement is the one that most sharply bounds the historiographic use of the catalogue. An absent cross-reference is not evidence of a missing record in the Archives, only of a missing entry in the release.

A natural question is whether the catalogue is unusual in this respect. We cannot answer the question in the terms the catalogue itself would recognise, since the Archives are not a comparable repository to any that is accessible to Republic scholarship. A comparison with the released catalogue of a comparable civilian repository (Royal Alderaanian Academy of Sciences, 32 BBY) returns a 91.8% resolution rate on a comparably designed sample, and the Alderaanian catalogue is subject to no analogue of the withholding mechanism that bears on the Jedi release. The ten-percentage-point difference is in the direction one would predict from the release structure alone and does not imply a difference in the Archives' own record-keeping.

The audit does not describe the full Jedi Archives. The keepers of the Archives have released what they have released, and the question of the full catalogue is beyond the released evidence; even the question of the ratio of released to withheld entries is one the released catalogue does not answer. We treat the keepers' practice as the horizon of the work, not as a subject of criticism.

6. Limitations of the Record

The audit samples one release of one catalogue for one fifty-year window. A later release of the same catalogue has retouched entry numbers in several blocks (Jedi Archives, 26 BBY release), and a comparable audit on the retouched release would yield a slightly different headline rate; we have not repeated the audit on it. The seven-field subject scheme is the catalogue's own, and some entries that would read as biographical in another scheme are classified here as philosophical commentary; the subject-field breakdown is therefore sensitive to the catalogue's indexing choices and not to any single external standard.

Three technical limitations deserve naming. The resolution test depends on the released summary text, which is itself an abridgement; a cross-reference whose context is thin may be judged ambiguous because the summary is thin, not because the Archives are. The clustering adjustment in the confidence interval treats each sampled entry as independent of every other; a release-wide error pattern (for example, a block of near-number errors that originated in a single scribe's work) would make the effective sample smaller than the clustered interval reports. And the ten-block period analysis treats the catalogue's internal dating as sound; the cross-reference audit does not test it.

Finally, we have not attempted to contact the keepers of the Archives to resolve the absent category. The released catalogue is the point of contact for Republic scholarship, and the question we have asked is a question about that catalogue. A different question, directed to the full Archives through a request procedure, is the keepers' to answer.