The Old Welsh Revitalisation Framework (OWRF) is a phonological regression and archaic restoration methodology for revitalising Old Welsh (c. 800–1150 CE) as a fully operational language state within the Brittonic Convergent Diachronic Revitalisation System (BCDRS). Where the MWRF operates from corpus abundance (~2.8 million words), the OWRF operates from severe corpus scarcity — the entire Old Welsh textual record amounts to only a few thousand words of continuous or semi-continuous prose.
The framework applies systematic phonological regression from the
Revitalised Middle Welsh (wlm) input column, using documented sound-change
correspondences between Old and Middle Welsh established by Strachan, Willis, and Jackson.
The O1 → O2 → O3 pipeline produces the revitalised Old Welsh
(owl) column which feeds into the
Northern Brittonic Toponymic Revitalisation Framework (NBTRF).
The system is governed by PBC's peer-review protocol.
The OWRF is the critical bridge in the BCDRS chain: it converts the relatively abundant Middle Welsh forms into their archaic Old Welsh antecedents, from which the NBTRF can then apply northern Brittonic divergence modelling. The methodological challenge of the OWRF is the reverse of the MWRF's: not selection among abundant evidence, but principled reconstruction from extreme scarcity, constrained by the hard discipline of working only backwards along documented sound-change paths.
Old Welsh — the earliest attested form of the Welsh language — was spoken and written between approximately 800 and 1150 CE. It represents the phase of the language immediately following the fragmentation of the Brittonic dialects of post-Roman Britain into recognisably distinct regional forms, and immediately preceding the classical Middle Welsh period with its rich literary and legal corpus. Old Welsh occupies a position of pivotal importance in the history of the Brittonic languages: it is the form from which both Middle Welsh and Cumbric directly descended, and its phonological and morphological structure is therefore the gateway through which the BCDRS pipeline passes from Middle Welsh to the northern Brittonic dialects.
The great challenge of Old Welsh linguistics is, simply, the scarcity of the surviving corpus. Unlike Middle Welsh — which offers 2.8 million words of digitised prose — the complete Old Welsh textual record amounts to only a few thousand words, preserved in forms that are fragmentary, mixed with Latin, or embedded in later manuscript traditions [1]. This is not the scarcity of a language that failed to produce literature; it is the scarcity of a language whose literature, if it existed, has not survived. The evidence that does survive is preserved almost entirely in contexts of scholarly and ecclesiastical Latin writing — as glosses in the margins of Latin manuscripts, as computational and calendrical notes, and as brief legal and memorial inscriptions.
The principal primary sources are:
These sources together provide a corpus adequate to establish the principal features of Old Welsh phonology, morphology, and syntax — and to support the O1 attestation tier for a small number of high-frequency grammatical forms. However, for the majority of the 307 dataset rows, no direct Old Welsh attestation is available. This makes the OWRF's O2 phonological regression stage the workhorse of the framework.
Old Welsh is linguistically close to Middle Welsh — closer, in most respects, than Middle Welsh is to Modern Welsh. The most significant differences are phonological and orthographic rather than structural: Old Welsh preserves several consonantal distinctions that were lost or merged in Middle Welsh, and its scribal conventions reflect Latin letter-habits that were gradually replaced by more distinctively Welsh orthographic practices as the Middle Welsh period developed. The OWRF's task is to systematically reverse those transitions — to work back from the Middle Welsh forms produced by the MWRF to their Old Welsh antecedents.
Phonological regression — the principled reconstruction of an earlier phonological form from a later one, using documented sound-change correspondences — is the established methodology of historical linguistics. It is what linguists mean by "reconstruction": not speculation about what a language might have sounded like, but the systematic reversal of known, documented, and independently verified sound changes [7].
The specific sound changes between Old Welsh and Middle Welsh are not hypothetical. They are documented in the major works of Celtic historical linguistics — above all in Kenneth Jackson's Language and History in Early Britain [7], which provides the most comprehensive account of Brittonic phonological history from proto-Celtic through to the medieval Welsh and Cornish periods. Jackson's work establishes, for each documented sound change, the approximate date, the phonological environment, the evidential basis, and the comparative Brittonic context. The OWRF's O2 rules are derived directly from Jackson's documented correspondences and from the more recent syntheses provided by Willis [8] and Sims-Williams [6].
The phonological regression methodology has a clear discipline: it may only apply documented rules in documented environments. It may not apply a rule speculatively in an environment where the rule is not established. And it must default to Middle Welsh baseline retention (O3) wherever the rule application is ambiguous, the environment is unclear, or the evidence for the sound change is insufficient. This discipline is the OWRF's analogue of the NBTRF's Uncertainty Rule.
The Computus Fragment — preserved in Cambridge MS Add. 4543 and dated to approximately 920 CE — is the primary attestation source for the OWRF [2]. It provides directly attested Old Welsh forms for several grammatical items of relevance to the dataset: verbal forms of bod, prepositional forms, pronouns, and a handful of temporal nouns. These attested forms yield Grade A outputs at the O1 stage and provide the closest thing to direct Old Welsh evidence available in the corpus.
The Computus Fragment is not merely a lexical source. Its orthographic conventions — the use of single ⟨d⟩ for /ð/, the use of ⟨p⟩ for /f/ in certain positions, the specific spelling of pronominal and prepositional forms — provide the authoritative model for the OWRF's orthographic normalisation. Where the Computus spells a form in a particular way, that spelling is adopted as the standard OWRF form for that item.
Strachan's edition of the Old Welsh texts, published in 1909 [1], remains the standard critical edition for these materials. Willis's more recent synthesis [8] provides updated linguistic analysis but does not supersede Strachan as an edition of the primary sources.
The Old Welsh corpus is too small to provide evidence for all the forms required by the 307-entry dataset. The OWRF therefore draws systematically on comparative Brittonic evidence — principally from early Cornish and Breton, which share the Old Welsh period as their common ancestor — to support O2 regression where the Old Welsh sources are silent [9, 10].
Koch's work on Brittonic comparative linguistics [11] and James's Brittonic Language in the Old North (BLITON) [12] provide additional evidence from the northern Brittonic record that is directly relevant to the OWRF's task. The northern Brittonic zone — the geographical origin of Cumbric — was conservative in its phonological development, which means that Old Welsh forms representing the general Brittonic state of the period are likely to be a reliable approximation of the northern forms that Cumbric inherited.
Beyond phonological regression, the OWRF applies archaic morphological restoration in selected paradigm cells — most significantly in the verbal system, where Strachan's paradigm tables document Old Welsh synthetic inflections that were subsequently lost or replaced by periphrastic constructions in Middle Welsh [1].
The most important case is the present tense 3sg of bod (to be): Middle Welsh yw corresponds to Old Welsh iss/is, the latter form being directly attested in the Computus and in manuscript glosses. Similarly, the Old Welsh verbal noun of the verb "to go" is mynet (Middle Welsh mynd) — a form attested in Old Welsh glosses and providing one of the clearest documented divergences between the two periods. These restoration cases are relatively few in number but are among the highest-confidence outputs the OWRF produces.
The OWRF operates a strict hierarchical evidence model. The fundamental asymmetry with the MWRF is that Grade A (direct attestation) is achievable for a much smaller proportion of entries — perhaps 5–15% of the dataset rather than the 55–65% achievable at M1 in the MWRF. The OWRF's default outcome is Grade B (phonological regression) rather than Grade A, and Grade C (Middle Welsh baseline retention) is more frequently applied than in the MWRF.
| Level | Source | Method | Grade |
|---|---|---|---|
| O1-Computus | Computus Fragment (Cambridge MS Add. 4543) | Direct attestation of Old Welsh form | A |
| O1-Lichfield | Lichfield Gospels (Surrexit memorandum) | Direct attestation; early formulaic and legal Welsh | A |
| O1-Juvencus | Juvencus poems | Direct attestation; OW verse forms | A |
| O1-Glosses | Old Welsh manuscript glosses, per Strachan/Willis | Direct attestation; lexical and grammatical items | A |
| O1-GPC | GPC historical entries for Old Welsh period forms | Dictionary attestation with primary source citations | A |
| O2-Jackson | Jackson, Language and History in Early Britain | Phonological regression per Jackson's documented correspondences | B |
| O2-Willis | Willis, Old and Middle Welsh | Phonological regression per Willis's updated analysis | B |
| O2-Strachan | Strachan, Introduction to Early Welsh | Paradigm regression per Strachan's morphological tables | B |
| O3-wlm | Revitalised Middle Welsh column | wlm value retained; OW and MW forms identical or indistinguishable | C |
A crucial point about the evidence hierarchy: the O3 baseline in the OWRF is the wlm value (Revitalised Middle Welsh), not the cy value (Modern Welsh). The MWRF has already performed the work of identifying the Middle Welsh form; the OWRF takes that form as its starting point. This means that where the wlm and owl values are identical (Grade C), they may nonetheless differ from the Modern Welsh cy value — the distinction between cy and wlm having been established by the MWRF.
The OWRF processes each dataset entry through a three-stage sequential pipeline. The output
of each stage becomes the input to the next, and the final output of the pipeline is the
value written to the owl (Old Welsh) column.
Sources: Computus Fragment, Lichfield Gospels, Juvencus poems, manuscript glosses (per Strachan and Willis), GPC historical entries for Old Welsh.
O1 asks: is this word or paradigm form directly attested in a surviving Old Welsh primary source? This is the most demanding question the OWRF can ask, and — given the sparseness of the Old Welsh corpus — it can be answered affirmatively for relatively few dataset entries. The items most likely to yield O1 Grade A outputs are: verbal forms of bod attested in the Computus; the verbal noun mynet attested in manuscript glosses; pronominal forms confirmed by the Lichfield Gospels; and a small number of common nouns and particles attested across multiple sources.
The O1 procedure involves four sequential checks:
Source: wlm (Revitalised Middle Welsh) value; Jackson, Willis, Strachan for rule basis.
O2 applies systematic phonological regression to the wlm form, transforming it into its Old Welsh antecedent by reversing the documented sound changes between Old and Middle Welsh. This is the workhorse stage of the OWRF — the stage that handles the large majority of entries for which O1 attestation is unavailable.
O2 outputs are assigned Grade B. They represent academically defensible derivations from Jackson's documented correspondences — not attested forms, but principled reconstructions. The key discipline of O2 is to apply regression rules only in the environments where they are established and to default to O3 wherever the rule application is ambiguous.
Source: wlm (Revitalised Middle Welsh) value, retained directly.
O3 is applied when O1 attestation is unavailable and O2 regression either produces the same result as the wlm form or cannot be reliably applied. In these cases, the OWRF retains the wlm value and assigns Grade C.
Grade C is applied more frequently in the OWRF than in the MWRF, because the Old Welsh corpus is far smaller. For many items — particularly in the stable categories (days of the week, months, conjunctions, greetings) — the Old Welsh and Middle Welsh forms are effectively identical, and the O2 rules do not apply. For these items, the wlm value is the best available approximation of the Old Welsh form, and it is adopted as Grade C.
The following categories are expected to yield Grade C outputs as their standard baseline in the OWRF. The basis for this classification differs subtly from the MWRF's stable categories: while the MWRF identifies stable categories on the grounds that MW and ModW forms are effectively identical, the OWRF identifies them on the grounds that OW and MW forms are effectively identical — which is independently true, reflecting the conservatism of these lexical categories across the Old and Middle Welsh period.
| Row prefix | Category | Basis for Grade C baseline |
|---|---|---|
DAY_* | Days of the week | Latin-derived borrowings stable across OW and MW; attested in Computus Fragment as identical or near-identical to MW forms |
MON_* | Months | Latin borrowings; stable across both periods |
SEA_* | Seasons | Gwanwyn, Haf, Hydref, Gaeaf — stable across OW and MW |
TIM_* | Telling the time | Post-medieval register; wlm = cy = owl for these items |
TMP_* | Temporal words | Stable; OW and MW forms identical or near-identical |
CONJ_* | Conjunctions | Principal Welsh conjunctions largely stable across both periods |
PREP_* | Prepositions (uninflected) | Simple prepositions stable between OW and MW; inflected forms may differ — check Strachan for specific items |
GRT_*, INT_*, POL_* | Greetings / introductions / politeness | Post-medieval phrasebook register; no OW equivalents; wlm = owl for these items |
The key contrast with the NBTRF's immutable categories: the OWRF has no immutable categories in the sense of structural prohibitions on investigation. Grade C is an expected evidentiary baseline, not a theoretical constraint. If O1 attestation provides a genuinely different Old Welsh form for a day of the week or a conjunction, that O1 form takes precedence. Grade C reflects the state of the evidence, not a rule about what can be discovered.
However, the OWRF applies Grade C more broadly than the MWRF because the OW corpus is so sparse that even for active derivation categories, the regression rules may not apply — leaving wlm retention as the only defensible option. The OWRF practitioner must be alert to this: a row that looks like an active derivation candidate (e.g., an adjective with ⟨dd⟩ in the wlm form) may still receive Grade C if the regression environment for that specific item is unclear.
The following categories are expected to yield Grade A or Grade B outputs through O1 attestation or O2 regression. For these categories, either the Old Welsh primary sources provide direct evidence, or the phonological regression rules apply clearly and produce a form distinguishable from the Middle Welsh baseline.
| Row prefix | Category | Expected grade | Method |
|---|---|---|---|
BE_* | bod paradigm | A/B | Strachan §§87–88; Computus attestation for some cells (3sg *iss*); O2 for remainder |
GO_* | mynd paradigm | A/B | OW verbal noun *mynet* attested in glosses (Grade A); finite forms per Strachan §§101–102 (Grade B) |
PRN_* | Pronouns | A/B | OW and MW pronouns largely identical; Lichfield and Computus confirm key forms (Grade A) |
ADJ_* | Adjectives with ⟨dd⟩ or ⟨ff⟩ | B | O2 Rules 4a/4b/7: ⟨dd⟩ → ⟨d⟩; ⟨ff⟩ → ⟨p⟩ or ⟨f⟩ |
HAVE_* | cael/caffael paradigm | B/C | O2 orthographic rules where applicable; wlm baseline for cells not in Strachan |
COME_* | dyfod paradigm | B/C | O2 orthographic/phonological rules; Willis provides some OW finite forms |
NUM_* | Cardinal numbers | B/C | OW number system largely MW-identical; GPC confirms where attested |
ORD_* | Ordinal numbers | B/C | OW ordinal endings per Strachan where documented |
All OWRF outputs carry one of three confidence grades:
| Grade | Definition | Typical basis | ATTESTATION_CLASS |
|---|---|---|---|
| A | Directly attested in an Old Welsh primary source (Computus, Lichfield, Juvencus, glosses) as documented by Strachan, Willis, or GPC | Computus verbal form; Lichfield pronominal form; GPC Old Welsh entry with manuscript citation | DIRECT_ATTESTATION |
| B | Systematically derived by O2 phonological regression from the wlm form, using Jackson's documented sound correspondences | ⟨dd⟩ → ⟨d⟩ orthographic conversion; /g/ → /ɣ/ soft mutation restoration; synthetic verb ending from Strachan | PHONOLOGICAL_REGRESSION |
| C | Middle Welsh baseline retained; OW form identical to wlm or insufficiently evidenced for regression | Stable category; O2 rules do not apply to this item; OW and MW forms confirmed identical by sources | WLM_BASELINE |
The grade distribution in the OWRF is expected to show a much higher proportion of Grade C than the MWRF, reflecting the sparseness of the Old Welsh corpus. Where the MWRF achieves Grade A for approximately 55–65% of entries through Evans and GPC, the OWRF achieves Grade A for perhaps 5–15% — primarily verb forms of bod and mynd, and a small number of pronominal and prepositional forms confirmed by the primary sources.
The critical implication for the downstream NBTRF: the NBTRF takes the owl column as its primary input. For the majority of owl cells, the value will be Grade B or C — a phonologically regressed or wlm-retained form rather than a directly attested Old Welsh form. This does not undermine the NBTRF's outputs; the identity principle (xcb = owl for most entries) means that the owl column's Grade B and C values are inherited into the xcb column with explicit confidence documentation, maintaining full traceability.
Section §8 is the most technically specific section of the OWRF dissertation. The rules below represent the documented sound-change correspondences between Middle Welsh and Old Welsh, applied in reverse (regression direction) from wlm to owl. Each rule is traceable to Jackson (1953), Willis (2009), or Strachan (1909).
This is the most diagnostically important difference between Old Welsh and Middle Welsh, and it is the rule with the most significant practical effect on the OWRF dataset.
In Modern Welsh and Middle Welsh, the soft mutation (lenition) of an initial /g/ produces zero — the /g/ is deleted entirely, leaving no consonant. This Ø outcome is uniform across both periods. In Old Welsh, however, the corresponding lenited consonant was /ɣ/ — the voiced velar fricative — the phoneme that stands between the /g/ of unlenited position and the zero of Middle Welsh lenited position [7, §§25–30].
The historical development is: Proto-Brittonic /g/ → OW lenited /ɣ/ → MW/ModW lenited Ø (deletion). The deletion of /ɣ/ is a Middle Welsh development. Old Welsh retained the fricative.
OWRF Application: Wherever a wlm form shows Ø (zero) in a position where soft mutation of underlying /g/ is triggered — after feminine noun determiners, certain prepositions, the predicative particle yn, in the second element of compounds following a vowel-final first element — the owl form must restore initial ⟨g⟩ representing /ɣ/.
Orthographic representation: The OWRF represents Old Welsh /ɣ/ as ⟨g⟩, following the convention of the primary sources (which do not use a distinct grapheme for /ɣ/). The distinction is contextual: underlying /g/ appears in unmutated positions; /ɣ/ appears in mutated positions. Modern scholarly editions typically mark this distinction with diacritics or subscript notation, but the OWRF adopts the ⟨g⟩ convention of the manuscripts.
| Middle Welsh form (wlm) | Old Welsh form (owl) | Notes |
|---|---|---|
| (y) las — mutated form of *glas* (blue/green) | (y) glas | OW retains /ɣ/ spelled ⟨g⟩ where MW has Ø; Jackson §28 |
| Unmutated *glas* (blue/green) | glas | Identity — unmutated /g/ unchanged in both periods |
| (y) wyrdd — mutated form of *gwyrdd* (green) | (y) gwyrdd / (y) guird | OW ⟨u⟩ for /w/; /ɣ/ retained; Jackson §28 |
Middle Welsh represents the voiced dental fricative /ð/ with the digraph ⟨dd⟩ — a distinctively Welsh spelling convention that emerged in the later Old Welsh / early Middle Welsh transition period. Old Welsh, following Latin scribal habits, represented /ð/ with single ⟨d⟩ [1, 8].
Application: Replace all wlm ⟨dd⟩ (representing /ð/) with ⟨d⟩ in the owl form. This is one of the most consistently applicable rules across the dataset — any wlm form containing ⟨dd⟩ requires this substitution.
| Middle Welsh (wlm) | Old Welsh (owl) | Notes |
|---|---|---|
| rhudd (red) | rhud | ⟨dd⟩ → ⟨d⟩; Strachan confirms OW orthography |
| oedd (was, 3sg impf.) | oed | ⟨dd⟩ → ⟨d⟩; Computus attests oed |
| mynd (to go) | mynet | Verbal noun form differs entirely — see morphological restoration |
Middle Welsh represents the voiceless labio-dental fricative /f/ with the digraph ⟨ff⟩. Old Welsh used ⟨p⟩ in many positions (following the Latin convention for the bilabial fricative derived from Latin F) or single ⟨f⟩ — but did not use ⟨ff⟩ [1]. The ⟨p⟩ convention is particularly evident in forms like OW cap for MW caff (may get) and in the verbal noun of the verb "to get": OW capaul alongside the MW form caffael.
Application: Replace wlm ⟨ff⟩ with ⟨p⟩ (dominant scribal convention) or ⟨f⟩ in the owl form. The choice between ⟨p⟩ and ⟨f⟩ depends on position and the specific item — check Strachan for individual cases. In word-initial position, ⟨f⟩ is more common; in medial and final positions, ⟨p⟩ appears frequently.
Both Old and Middle Welsh represent /k/ with ⟨c⟩. This is an identity rule — no modification required. The use of ⟨c⟩ for /k/ is a conservative Latin-derived convention retained across both periods.
In some Old Welsh scribal sources, the voiceless dental fricative /θ/ (Middle Welsh ⟨th⟩) is represented by ⟨t⟩, particularly in word-final position. This is inconsistently applied across Old Welsh manuscripts and reflects the Latin scribal convention of using ⟨t⟩ for dental sounds. The OWRF applies this rule with caution: only in cases where Strachan or Willis specifically documents the ⟨t⟩ spelling for a particular item.
Old Welsh and Middle Welsh pronoun forms are largely identical, making this one of the most conservative areas of the grammatical system. The OWRF applies identity for most pronominal rows. The 3pl demonstrates slight variation: both hwy and wynt are found in Old Welsh sources; the OWRF uses hwy as the standard form, following the wlm baseline set by the MWRF.
| Person | Middle Welsh (wlm) | Old Welsh (owl) | Notes |
|---|---|---|---|
| 1sg | mi | mi | Identity; Lichfield confirms |
| 2sg | ti | ti | Identity |
| 3sg masc | ef | ef | Identity |
| 3sg fem | hi | hi | Identity |
| 1pl | ni | ni | Identity |
| 2pl | chwi | chwi | Identity |
| 3pl | hwy | hwy / wynt | Both attested in OW; hwy used as standard |
Old Welsh preserves more synthetic verbal inflections than Middle Welsh, particularly in the paradigm of bod (to be). Key differences documented in Strachan's tables:
| Tense / Person | Middle Welsh (wlm) | Old Welsh (owl) | Source |
|---|---|---|---|
| Present 3sg | yw | iss / is | Strachan §87; Computus attestation (Grade A) |
| Present 1sg | wyf | bim / bof | Strachan §87 (Grade B) |
| Imperfect 3sg | oedd | oed | Strachan §88; glosses (Grade A) |
| Verbal noun (to go) | mynd | mynet | Glosses; Jackson §105 (Grade A) |
The Old Welsh definite article y/yr is identical to the Middle Welsh form. The major prepositions are largely identical between Old and Middle Welsh. Where Strachan documents a divergent Old Welsh prepositional form, it is applied; otherwise, the OWRF applies identity for prepositions (wlm = owl).
The OWRF pipeline has been applied to all 307 entries in the Revitalised Cumbric dataset. The distribution of outcomes reflects the evidentiary conditions of the Old Welsh corpus: much higher Grade C proportions than the MWRF, with Grade A confined to a small number of high-frequency grammatical forms.
| Stage | Expected proportion | Notes |
|---|---|---|
| O1 — Grade A (Direct Attestation) | ~5–15% of entries | Verb forms of bod (Computus); verbal noun mynet; key pronominal and prepositional forms |
| O2 — Grade B (Phonological Regression) | ~30–45% of entries | Adjectives with ⟨dd⟩/⟨ff⟩; verb paradigm cells in Strachan; items where /g/ → /ɣ/ rule applies |
| O3 — Grade C (wlm Baseline) | ~45–60% of entries | Stable categories; items where OW and MW are identical; regression rules do not apply |
The high Grade C proportion in the OWRF is not a weakness of the framework — it is the correct scholarly outcome given the evidentiary conditions. The OWRF never fabricates a Grade A or B outcome where the evidence does not support it. A Grade C entry (owl = wlm) is a high-confidence claim: it asserts that the Old Welsh form is the same as the Middle Welsh form for this item, which is itself a substantive scholarly judgement grounded in the evidence hierarchy.
| Domain | Expected owl/wlm relationship | Rule applied |
|---|---|---|
| bod present 3sg | Divergent: owl iss, wlm yw | O1 — Computus attestation (Grade A) |
| mynd verbal noun | Divergent: owl mynet, wlm mynd | O1 — gloss attestation (Grade A) |
| Adjectives with ⟨dd⟩ | Divergent: owl ⟨d⟩, wlm ⟨dd⟩ | O2 Rule 2 (Grade B) |
| Forms with ⟨ff⟩ | Divergent: owl ⟨p⟩ or ⟨f⟩, wlm ⟨ff⟩ | O2 Rule 3 (Grade B) |
| Soft-mutated /g/ forms | Divergent: owl initial ⟨g⟩, wlm Ø | O2 Rule 1 (Grade B) |
| Days, months, seasons | Identity: owl = wlm | O3 (Grade C) |
| Most pronouns | Identity: owl = wlm | O1 (Grade A); forms identical |
The following limitations apply to all OWRF outputs:
The OWRF occupies the middle position in the medial pipeline of the Brittonic Convergent Diachronic Revitalisation System (BCDRS). It receives input from the MWRF and passes its output to the NBTRF. This position is methodologically pivotal: the OWRF is the bridge between the period of Welsh linguistic abundance and the period of near-total Cumbric absence.
Modern Welsh (cy) → MWRF → Revitalised Middle Welsh (wlm) → OWRF → Revitalised Old Welsh (owl) → NBTRF → Revitalised Cumbric (xcb)
The OWRF's role in the BCDRS is the conversion operation: it takes the Middle Welsh forms produced by the MWRF and transforms them into Old Welsh antecedents. This matters for the NBTRF because Cumbric was not a dialect of Middle Welsh — it was a dialect of Old Welsh and Late Brittonic. The relationship between Cumbric and the southern Brittonic tradition runs through the Old Welsh period, not the Middle Welsh period. Without the OWRF, the NBTRF would be deriving Cumbric from Middle Welsh forms — forms that already reflect sound changes (the loss of /ɣ/, the development of ⟨dd⟩ and ⟨ff⟩ digraphs, and other Middle Welsh innovations) that post-date the Cumbric period. The OWRF removes these Middle Welsh innovations, restoring the Old Welsh baseline that the NBTRF can legitimately use as the Cumbric starting point.
In one important sense, the OWRF is the most theoretically significant of the three BCDRS frameworks. The MWRF operates in conditions of abundance — its primary challenge is selection. The NBTRF operates under extreme evidentiary constraint — its primary challenge is the discipline of not speculating. The OWRF operates at the inflection point between these two conditions: it applies the tools of historical linguistics to move from abundance to reconstruction, using documented sound changes as the only legitimate path.
The relationship between the OWRF and the NBTRF is one of strict dependency. The NBTRF treats the owl column as its authoritative baseline for the 282 identity entries (xcb = owl). Any improvement to the owl column — any correction of a Grade C entry to Grade B based on new evidence, any correction of a Grade B entry to Grade A based on new attestation — flows automatically through to the xcb column. The integrated nature of the BCDRS pipeline means that improvements anywhere in the chain improve the final output.
For the Cumbric revitalisation project specifically, the OWRF's most significant contribution is the restoration of the /g/ → /ɣ/ distinction in soft mutation environments. This distinction is invisible in the Middle Welsh baseline but is real in Old Welsh — and since Cumbric descends from Old Welsh, the distinction was present in Cumbric in the pre-decay period. Whether the NBTRF should reproduce the /ɣ/ in the xcb column is a question for the NBTRF specification — but the OWRF ensures that the evidence for its existence is properly represented in the owl baseline.
The OWRF is designed as a living framework. Old Welsh scholarship, though a specialised field, is active: new critical editions of Old Welsh texts, new analyses of manuscript glosses, advances in comparative Brittonic phonological history, and the growing digitisation of medieval manuscripts all offer prospects for improving the quality of the owl column over time.
Sims-Williams's ongoing work on early Welsh manuscripts [6] is particularly relevant to the OWRF's O1 attestation tier — his systematic cataloguing of Old Welsh manuscript glosses provides the most comprehensive contemporary survey of the primary sources. New editions of early Welsh texts produced by the University of Wales Centre for Advanced Welsh and Celtic Studies continue to expand the edited corpus.
The most significant future improvement to the OWRF would be the expansion of the O1 Grade A tier through new attestation discoveries. Each new Old Welsh gloss or text fragment has the potential to confirm or refine an owl value currently assigned Grade B or C. The framework is designed to accommodate such improvements without requiring structural change — a new O1 finding simply updates the grade and source for the relevant entry in the trace file and the dataset.
External contributions from qualified specialists in Old Welsh linguistics are welcomed under the formal contribution protocol documented at the Contributors page. Contributions relating to the OWRF are held to the same standard as NBTRF contributions: dissertation-format submissions, explicit evidence documentation, and graded confidence assignments. The field expertise required is particularly narrow — the OWRF sits at the intersection of Old Welsh philology and Brittonic historical phonology — and specialist review is especially important for this framework.
The conservative standard is permanent throughout the BCDRS. The OWRF never fabricates Grade A or B outputs where the evidence does not exist. Each Grade C entry is a considered scholarly position, not a gap to be filled by speculation. The goal is not to minimise the number of Grade C entries — it is to ensure that every grade assignment is accurate, every source is documented, and every correction is traceable.