Changelog

Notable changes to DERBi PIE, most recent first.

Currently running build 9255b67 · 10/25/2026, 4:31 PM ET

2026-10-25 — Mallory & Adams matching; notes on sources; Old Irish and OCS repaired
  • About eighty more entries from Mallory & Adams’ Oxford Introduction (from men- to reudʰ-) are matched to their roots, by Tim Dempsey. Where they have no counterpart, they are new roots, e.g. men- ‘chin’, menk- ‘lack’, perd- ‘panther’, pēnt- ‘heel’.
  • A source can now carry a short note and a link to another root, shown under its line at the top of a root page. Mallory & Adams’ *pḗr ‘house’ is a new root, marked uncertain since they doubt it is Proto-Indo-European; its note links it to *prihₓós ‘one’s own, dear’ under prei̯H- ‘to become dear’, with which they tentatively connect it. The two roots themselves are not linked.
  • mono- ‘neck’ is now listed under men- ‘to tower’, with mono/i- ‘necklace’ under mono-.
  • Pokorny’s per- ‘forward, through’ (2a) is labelled by his own subdivisions a.–i. (per, peri; pero-s; prai; peres; pr̥-; peră; pro, prō; preti, proti; porsō(d)). His separate entry prō̆- is only a cross-reference to per- and now points there.
  • Where one source cites several entries for a root without forms on the page, they are now shown on one line.
  • Fixed: after this morning’s update, Old Irish and Old Church Slavonic dictionary entries briefly showed garbled letters (accented and Cyrillic characters), and some Old Irish forms linked to the wrong entry. Both are repaired.
2026-10-25 — More of Pokorny’s entries separated by root
  • Twenty-two more of Pokorny’s entries that print two root shapes in one headword are now divided, this time form by form: for example 2. ĝheu- ‘to disappear’ beside ĝheu-d-, 1. kreu- ‘raw flesh, blood’ beside kreus- ‘ice, crust’, 2. tu̯ei- beside tu̯ei-s-, and mē- beside m-e-t- in ‘to mow’ and ‘to measure’. Each part is filed under its own root.
  • A form goes with the longer shape when it shows the extra sound (the -d- of Lithuanian žudýti ‘to kill’ beside žū́ti ‘to perish’); forms that Pokorny himself explains as a suffix formation stay with the shorter root.
  • Nineteen roots are added in Pokorny’s shape where LIV and NIL have no counterpart, for example kreu̯s- ‘ice, crust’, perg- ‘to hit’ and ukʷs- ‘to wet’, from which Pokorny derives the word for ‘ox’.
  • A section can now carry a short note in the “show all” table. Pokorny gives only sler- “with various enlargements” for Latvian slarpata ‘rag’, German schlarfen ‘to shuffle’ and similar words; they are now filed under slerp- and slerg-, with a note that these two shapes are not Pokorny’s own.
  • Where Pokorny offers a whole group only with “vielleicht” (perhaps), its forms are now shown as uncertain (e.g. skeu-k- for English shy).
2026-10-24 — Pokorny’s entries separated by root
  • Many of Pokorny’s entries combine several roots under one headword (for example 9. au̯(e)- with au̯ed- ‘water’ and au̯er- ‘water, rain’, or 2. i̯eu- ‘to join’ with i̯eu̯-g- ‘to yoke’). About 155 such entries are now divided into Pokorny’s own sections, and each section is filed under its own root: au̯ed- under u̯ed-, the yoke words under i̯eu̯g-, gʷem- ‘to come’ under gʷem- rather than gʷeh₂- ‘to walk’.
  • On a root page, a Pokorny section that belongs to a larger entry shows a short line under its heading (“IEW 9 au̯(e)- § b · one of 4 sections”), with “show all” to see every section of the entry and the root it is filed under.
  • Pokorny’s reconstructions are shown in the same notation as the other sources (dʰeu̯dʰ- for his dheu-dh-); hover over a form to see it as printed.
  • Where Pokorny’s sections have no counterpart in LIV or NIL, they are listed as roots of their own in his shape (about 130 roots, e.g. dʰeu̯bʰ-, gneu̯bʰ-, reu̯b-).
  • Formations within an entry (Pokorny’s Verbum, Nominalbildungen, Formantien) stay with their root and are labelled.
2026-10-24 — Pokorny forms back under the right entries; lexicon links repaired
  • About 300 Pokorny forms that had been filed under a neighbouring entry are back where they belong. For example, the forms of 2. ĝen- ‘to know’ (jānā́mi, kunnan, γνῶμα) had been showing under 2. en- ‘year’, and the forms of 3. and 4. ei- (yew, rowan; pole, oar) under 1. ei- ‘to go’.
  • Pokorny’s se- (reflexive pronoun) is now under s(u̯)e ‘self’, not under set- ‘to be good’.
  • The root shown as pleu̯k- ‘to fly’ with only ‘pluck, feather, fleece’ forms is now pleu̯s- ‘to pluck’, as in Pokorny; three Latvian glosses on it were corrected.
  • About 200 links from cited forms to the Sanskrit, Greek and Latin lexicons were checked and repaired: forms now link to their own dictionary entry (βελτίων, κτείνω, pitṛ́, div, samudrá …), and links that pointed to an unrelated word were removed.
  • A few Greek, Avestan and Persian forms that were labelled Sanskrit now carry the right language.
  • LIV’s perfect of *h₁eu̯- ‘to see’ (Hittite ūhhi, autti, uwanzi) was missing and has been added; Vedic uvā is corrected to uvé.
2026-10-23 — New index: roots, languages, word classes, stem classes
  • The Index has four views. Roots: alphabetical, with its own sections for h₁ h₂ h₃ H, the aspirates, palatals and labiovelars; a jump bar; a forgiving filter (h2ek finds h₂ek̑-, bher finds bʰer-); and each root’s formations, variants and enlargements one click away.
  • Languages: every language with the number of forms cited for it, by branch, each language under its ancestor (Latin › Late Latin › Medieval Latin; Vulgar Latin › Romance › French …). A drawn family tree, after Ringe et al. as on the root pages, can be explored level by level; line thickness and dot size show how many forms come through. Searching zooms the tree to the language.
  • Word class and Stem class: roots and formations by part of speech and by formation type (root aorists, nasal presents, s-stems …), each with a filterable list.
  • Fragmentary Indo-European languages (Thracian, Illyrian, Messapic …) and non-Indo-European languages (Finnish, Basque, Etruscan …) are now listed separately instead of under “Other”.
  • Persian forms were listed under Anatolian; now Iranian. Cuneiform and Hieroglyphic Luvian are grouped under Luvian. A Welsh and Cornish form cited together by Pokorny is split into its two languages.
  • Removed leftover LIV formatting marks (##, footnote numbers) from a few reconstructions.
  • Links in the dark header of root pages are easier to read, and a root page that failed to open works again.
  • Corrected the credits: Tim Dempsey processed the Pokorny and NIL data.
2026-10-23 — Indo-Iranian branch + duplicate NIL forms removed
  • New branch Indo-Iranian for languages that are neither Indic nor Iranian: the Nuristani languages (Waigali, Kati, Ashkun, Prasun) and Proto-Indo-Iranian forms. Previously under “Other”.
  • Mitanni Indo-Aryan forms (personal names and terms preserved in Hurrian and Akkadian texts) now appear under Indic instead of “Other”.
  • Several NIL forms were shown twice, under two roots of the same shape: the words for “ear” and “dawn” (*h₂au̯s-os-, *h₂us-i-), “sow” and “give birth” (*suH-ko-), and the two roots *masd- “be fat” and “pole, mast”. Each form now appears only under the root NIL gives it.
  • NIL’s forms for “yesterday” now appear once, under *dei̯- “shine” as in NIL; the “yesterday” root has a note linking there.
  • Small corrections: Cornish auon “river” (was Old Cornish), Waigali caräy- “herd cattle” (the language name had become part of the form), Mitanni -za-na “people” (had been replaced by an Oscan form).
2026-10-23 — Greek and German meanings translated
  • About 480 English meanings that were still in Greek — mostly the ancient glosses of Hesychius that Pokorny and NIL quote (e.g. φαγόνες “jaws”, ἀβροῦτες “eyebrows (Macedonian)”) — are now translated into English. The Greek stays in the original-language gloss.
  • Forms that Pokorny glosses with the German prefix ver- (Gothic faír-, fra-, Old High German fir- …, under *per-) now have an English explanation.
  • Corrected a few source glosses that had been copied from the wrong line (Cypriot ti-we-i-te-mi-se, Greek ἀμέλδειν) and repaired several garbled ones.
2026-10-23 — Missing Pokorny and NIL forms restored
  • About 1,250 forms from Pokorny (IEW) that were missing from root pages are back. Most are forms IEW cites for several languages at once (e.g. Old Saxon and Old High German nio “never”); each language now has its own row.
  • NIL’s forms for “father-in-law” and “mother-in-law” (*su̯ék̑uro-, *su̯ek̑rúH-) and “fist” were missing and now appear.
  • Fixed IEW forms filed under the wrong root: the words for “naked” (Latin nūdus, Gothic naqaþs …) appeared under *no- “that” instead of *nogʷ-, and Greek ἀμέργω “strip off” etc. under *meregh- “soak” instead of *merg̑-.
  • About 140 forms without a language now have one — e.g. Latin minae, mōns, mentum under *men- — and forms that NIL marks as doubtful are flagged as uncertain.
  • Old Frisian forms (abbreviated “afr.”) were labelled Old French; corrected.
  • “Macedonian” in IEW and NIL is Ancient Macedonian and is now listed as such under Hellenic. Added Slovincian, Luxembourgish, Manx and Wanetsi (Waneci).
  • Fixed some forms that could not be found by searching for their spelling.
2026-10-23 — Language labels fixed
  • About 1,700 cited forms were shown under “Other” instead of their language and branch — among them most Czech, Danish, Aeolic Greek, Old Saxon and Tocharian B forms from Pokorny. They now appear under the right language, on root pages and in search results, and the branch and language filters find them.
  • Language names that had lost their accented letters (e.g. “??ech.” for Czech, “??ol.” for Aeolic) are restored throughout.
  • The original LIV² text and footnotes shown with each form had lost their special characters (ʰ, ā, ‘…’); restored from the source.
  • Forms that a source marks as doubtful by a “?” before the language are now consistently marked uncertain (hidden until you click “Show uncertain”). 46 NIL forms that had lost this mark have it back.
  • Added Wanetsi (Waneci), an Iranian language cited in NIL.
  • Removed a stray Old Saxon thunni “thin” wrongly listed under *su̯ep- “sleep”.
2026-10-22 — Pattern search repaired + choose roots, stems/words or reflexes
  • The search bar now lets you tick what to search for — Roots, Stems / Words, Reflexes, in any combination — and whether to match forms, English meanings, or both. Results are grouped under the same three headings.
  • Cover-symbol and feature patterns work again across the board: plain CVC, HVC etc. (previously treated as ordinary text), whole-root anchors like #CVC# (which missed every root written with a final hyphen), and palatal roots (k̑, ĝ) that the patterns skipped.
  • New and revised symbols: P = any plosive (stop); K = velar/palatal stop; R = resonant (r l m n w y). Hover over a symbol in the search box or on a root page for its meaning.
  • Patterns can name optional segments: k(')er = ḱer or ker, k(w)el = kel or kʷel, g(w)h = gʰ or gʷʰ; kw, gh, h2 can be typed for kʷ, gʰ, h₂. Alternatives can be joined with | as well as OR (#CRVC# | #CVRC#).
  • Patterns now also search reflexes, compared in plain spelling (Skt. kṛt matches #CRC#).
  • Every cover symbol in the search box is now highlighted and explained, not just the first.
2026-10-22 — Search by source + cleaner root pages
  • The search page now has three tabs: All References, Etymological Dictionaries and Language Dictionaries (the daughter-language dictionaries, formerly “Dictionaries”).
  • Under Etymological Dictionaries you can limit a search to LIV², NIL, IEW and/or Mallory & Adams. Results then show only what those sources contain — their roots, stems and cited forms — and root pages you open from the results stay filtered to them. “About these sources” lists each work with the number of roots it covers.
  • Source references on root pages are now plain small-capital sigla (liv² · iew · nil) in a column beside each form, instead of coloured labels.
  • Root pages show the part of speech and, for stems and derivatives, the formation type (e.g. “noun · ti-stem”, “verb · root aorist”). NIL stems whose forms carry a gender are now marked as nouns.
  • NIL and LIV root entries now show each work’s own reconstruction rather than a placeholder (“noun”, “verb”).
  • Root enlargements (e.g. *bʰeru̯- “boil” beside *bʰer-) are now separate entries linked from their base root, instead of having their forms merged into it. Compounds with two source roots now link to both.
  • Fixed LIV² citations missing their author and title.
  • Root pages can carry a short note; e.g. *abō(n) “ape” now notes the competing views, and Kroonen’s Etymological Dictionary of Proto-Germanic has been added as a reference.
2026-10-21 — Smarter search + LIV-style notation
  • Root search is much more forgiving: type plain ASCII (bhreg, gwem, kwetwores) and it finds bʰreg(ˆ)-, gʷem-, kʷetu̯er-. A bare h or H finds roots with any laryngeal; h after b, d, g is read as aspiration first.
  • Root search also copes with a dropped laryngeal (aug → h₂eu̯g-), a different ablaut grade, Pokorny spellings, a palatal typed as a plain k or g, and a typed ending (agros → h₂eĝ-).
  • h₂/₃ roots now turn up when searching for either h₂ or h₃.
  • Results are ranked by how closely they match, and each result says what it matched (e.g. “matched agro- (IEW)”).
  • When a search hits a stem or derivative that has its own forms, the result leads to that page (agro → *h₂aĝ-ro- “field”, listed under h₂eĝ-).
  • Reflex search works without accents or special characters, and finds Greek, Cyrillic and Devanagari words from a Latin-letter spelling.
  • Meaning search ranks roots whose gloss is exactly the word ahead of those that merely contain it; searching with quotation marks (“to be”) now works.
  • Related-meaning suggestions now agree between the All and Meaning tabs (pie → apple, broth), and only suggest roots that have daughter-language forms.
  • Stems and derivatives listed on a root page (LIV verb stems, NIL nominals) now link to their own pages, and those pages link back to the root.
  • Pages for roots with no daughter-language forms yet now say so, and name the source that cites them, instead of appearing blank.
  • Reconstructions are displayed LIV-style: optional palatal as k/g with a raised (ˆ), optional (ʷ) and (ʰ), h₂/₃, ā̆, and thematic e/o. Hover for a short explanation.
  • Cover symbols (T, K, P, R, L, N) are highlighted with a hover note, on root pages and as you type in the search box, along with a “Reads as” line explaining the query.
  • Fixed some Latin words wrongly showing Sanskrit forms (e.g. flamen), and headwords containing a slash being split in two.
2026-10-21 — Root data cleanup
  • Resolved a dozen pairs of roots that were each marked as a variant of the other. Each pair now has a single headword (e.g. *dh₂ék̑ru “tear”, *pleh₂k- “strike”, *teng(ˆ)- “think”), or both are kept separate where the sources treat them as distinct (IEW’s u̯edh- “lead” vs. “bind”; “woe” vs. onomatopoeic u̯ī̆-).
  • LIV’s *ster- “knock down” and *sterh₃- “spread out” are now separate roots, and IEW’s (s)ter- “stiff” has its own entry.
  • *k̑u̯ei̯t- “light up brightly” and LIV’s Balto-Slavic *?ku̯ei̯t- “glow” are now separate entries, linked as variants.
  • “Yesterday” (*dʰg̑ʰi̯és) is now listed under *dei̯- “shine”, following NIL (particle *ĝʰe + genitive *di̯és “of the day”).
  • “Ape” and “rye” now use Pokorny’s headwords (abō(n), u̯rughi̯o-); an unsourced derivation of “ape” from *h₂ep- “water” was removed.
  • LIV stem labels such as “R(e)-e-present” and “R(e)-i̯e-present” were showing as just “e”; they now display in full.
  • Removed duplicate root and source entries that made some reflexes appear twice.
2026-10-19 — Language list cleanup
  • Removed duplicate languages from the “Filter by branch” and “Filter by language” lists: where several source abbreviations meant the same language, they now appear once.
  • Merged duplicate variants such as Shkodran Albanian / Albanian (Shkodër), and consolidated Dutch, Low German, Sorbian, Swedish and Middle High German labels.
  • Fixed misspelled or garbled language names (e.g. “Alabanian” → Albanian, and broken accents in Aukštaitian and Taittirīya-Saṃhitā).
  • Southern Balochi is kept as its own variety, now listed as “Balochi (Southern)”.
2026-08-08 — Old Church Slavonic + Old Irish added
  • Added Old Church Slavonic and Old Irish dictionary entries (Kaikki-sourced), roughly 4,500 and 6,500 words respectively.
  • Linked around 750 Old Irish reflexes to their dictionary entries, so they now show up as clickable links under the relevant PIE roots.
  • Cleaned up a batch of duplicate Old Church Slavonic entries left over from the import process.
2026-08-02 — Dictionary data-quality pass
  • Corrected roughly 250 dictionary entries where a data-processing bug had corrupted the English translation text.
  • Filled in around 530 entries that were showing a broken "?????" placeholder instead of a real English translation.
  • Fixed spelling errors in about 30 verb translations (e.g. "shakeed" → "shakes").
2026-08-01 — Dictionaries expansion + meaning search
  • Added Old English, Lithuanian, and Old Norse to the Dictionaries tab.
  • Added a word/meaning toggle to the dictionary search bar, so a query like "fero" (word) and "carry" (meaning) can be told apart.
  • Fixed Latin meaning search not showing the English gloss under results.
  • Fixed "exact" match mode in meaning search returning no results for words with multi-sense glosses (e.g. searching "sky" against a gloss of "sky, heavens").
  • Root and stem entries sourced from LIV² and NIL that weren’t appearing in search results now show up correctly.
  • Fixed search sometimes missing results due to accent-mark mismatches between a typed query and stored data.
  • Related word forms under the same root are now grouped together as "Related stems" instead of showing as separate entries.