📄 Old English theru and Modern English Tharf-Cakes

Summary

Biggam's short philological note (SELIM 5; the article itself is dated 1996) addresses the obscure word þeru. It appears in King Offa's grant of Westbury-on-Trym (793–6), in a list of goods owed to the king. Andrew Breeze had already argued from Welsh and Breton parallels that the word means "loaves". Biggam extends his argument by proposing that þeru is related to OE þeorf "unleavened (bread)", which survives as Modern English tharf-cake.

The obstacle is the missing -f. She explains it through two observations:

• Old English often simplified clusters of three consonants in speech.
• The related verb *þurfan ("need, be obliged") lost its -f in Middle English.

She suggests the reduced form may reflect everyday speech among estate scribes. She also accounts for why unleavened bread could be high-status then but disparaged later: the word combines "obligation/dues" with "need/poverty", and Biblical tradition gave unleavened bread a sense of purity. Finally, she proposes a second unreduced-f example in the Old English Leviticus.

Key points

• The sole witness for þeru is Cotton Tiberius A xiii (Hemming's cartulary), dated to around 1016–1025. The phrase is "VI lang þero", meaning six long loaves.
• The plural ending implies a short-stemmed neuter noun with a nominative singular *þer, reduced from þerf.
• The earliest recorded -f loss in tharf/thar is c. 1290 (OED). Campbell's grammar, however, documents Old English cluster reduction, including þorleas from þorfleas.
• In Leviticus 8:31, Biggam reads the manuscript form þæra (in both Cotton Claudius B iv and Laud Misc. 509) as a partitive genitive meaning "unleavened loaves". This rejects Crawford's emendation to þær þa ("there then").
• Dialect evidence from the EDD shows that tharf later acquired negative senses such as "lumpish" and "unsociable".

Interesting ideas

• I1. A puzzling form is explained by a documented phonological process (cluster reduction) attested in a related word, rather than by an ad hoc story.
• I2. The grammatical number and stem class of an unknown noun are inferred from its context (a numeral) together with its inflectional ending.
• I3. A genitive object of a verb of eating is read as partitive, meaning "eat some of", which changes how the object relates to the event.
• I4. The author challenges an editor's "correction" and shows the original reading makes sense under a better hypothesis.
• I5. Two seemingly unrelated senses (obligation, flatness) are linked under one superordinate concept (pressing weight), which also explains an apparent contradiction in social status across periods.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tokens): The lemmatizer could map unknown variant spellings to known lemmas through a small set of attested reduction and variation rules, logging which rule fired so the mapping stays explainable.
• I2 (tokens): When a noun is out of vocabulary, morphology could combine syntactic cues such as numerals and determiners with suffix evidence to guess number and lemma instead of tagging it blindly.
• I3 (frames): Role assignment could flag partitive objects of consumption verbs so that the world model records partial rather than total consumption.
• I4 (learn): Self-induced repair rules should be checked against the unrepaired input, keeping the original whenever a coherent analysis exists, to avoid over-correcting silver data.
• I5 (world): The ontology could store polysemy as senses hanging off a shared abstract feature, which would help explain metaphorical uses and shifts in evaluative connotation.