Monthly Archives: May 2013

Let them eat *turta!

An etymological investigation into the tarts and torts concerning Proto-Romance *turta “loaf, piece of bread”

Roman bread

In Old French tarte “pastry” and torte “round bread” stand next to eachother from the 13th c. CE onwards. Probably via French these terms made their way into Modern English as tart, into Modern Dutch as taart (< MidDu. tarte) and into Modern German as torte[1]. Despite their tastiness the words have proven notoriously difficult to etymologize (see REW 8802 and 8890, and Bloch 1932: 333-34).

Both lexemes are often[2] derived from Rom. *torta pane “rolled or bent bread” (cf. Lat. torquēre “to twist, to roll, to bend”) as found in the vulgata translation of the bible (ca. 420 CE) where torta panis  glosses Gk. ἄρτος “bread” (Exodus 29 : 23, I Chronic. 16 : 3, Jeremias 37 : 21).  However, the sentences in the vulgata always read tortam panis or torta panis, in which panis is a genitive to torta, probably in order to render the meaning “a loaf of bread”. If the bread itself was bent or rolled we would expect torta to be an adjective agreeing with the noun. Since Latin panis (gen. panis) is neuter and all the reflexes of Rom. *torta in the daughter languages are masculine we cannot account for the feminine gender of *torta. Furthermore, phonologically a connection to Lat. torquēre does not fit since all languages seem to point to PRom. *turta which is clear from the cognates. In the East-Romance dialects we find Rum. turtă “cake,” Vegliot. turta “vier aneinander hängende kleine Broten,” who continue PRom. *turta unchanged. PRom. *turta regularly developed into Western Romance *torta as found in OFr. torte, tourte “pain de forme ronde” (cf. ModFr. tourte “pastry” [turt]) and ModIt. torta “cake” (see REW 8802). A derivate from Lat. torquēre on the other hand would have yielded WRom. *tɔrta (cp. ModFr. tordre [tɔrdrə] “to twist,” ModFr. tort “blame, mistake” [tɔrt]). Therefore it is clear that OFr. torte, tourte and its cognates in the other Romance sister dialects cannot be derived from Lat. torquēre.

Also the earliest attestation of the word on the 2nd c. CE Vindolanda tablets reads turtas (Tab.Vind. II 120.80,[3] for the interpretation see Adams 2009: 611). After that we encounter the word in the Byzantine chronicle of Theophanes (ca. 320 CE) as τουρτίον, pl. τουρτία “loaf of bread” (see Matthews 2009: 191). The ου spelling shows that the donor word had [u] as root vocalism at the time of the loan. A few centuries later on the continent we find turtam glossing collyridam “pastry roll, cake” in the 9th c. Reichenau glosses. Here the u is in all likelihood an orthographical representation for what must have been [o] already. Since the Reichenau glosses are very informative in relating to us lexical items from the colloquial registers of Romance (which in this period had already evolved into Old French) we may safely see in it a confirmation that the word turta/torta was commonly used in 9th c. Francia.

Two questions arise. What is the origin of PRom. *turta and how do we explain the variants that must go back to Rom. *tarta? To my mind we should allow the possibilty that the word does not have its origins in Latin, which would explain its relatively late attestation.

Since the word is first attested on the British isles it seems more than fair to start our search there. In Middle Welsh we find the word torth “loaf of bread, hump of bread” (see Bevon and Donovon 2001: 6014[4]) which has often been taken as a loan from Rom. *torta. However, the word might be connected to Middle Irish tort, toirt “heap, mass” which stands alongside Middle Irish tort, gen. torte f. “loaf of bread” (Bondarenko 2012: column 262[5]). All these forms can be traced back to Insular Celtic *turtā. Both in Welsh and Irish the *u would get raised to *o under influence of the following *ā. We should note that Middle Welsh quite regularly shifts *o before resonant + consonant clusters to *a in Latin loanwords, e.g. MidW parchell “piglet” < Lat. porcellus and MidW carrai “strap” < BritRom. *kɔrria < Lat. *corrigia (Morris Jones 1913: 87). If the word was really a loan from Latin or British Romance we would have expected a variant Middle Welsh **tarth next to torth. This strengthens our case for a Celtic origin for PRom. *turta.

Insular Celtic *turtā could then have entered British Romance from whence it may have spread all over the Romance dialect continuum. It is also possible a Gaulish cognate *turtā, formally identical to the Insular Celtic proto-forms, provided the donor word. In either way the semantics would then have moved from “hump, piece” to “hump of bread” and then finally “bread.” This development is paralleled in Germanic where OE bread “brit, crumb, morsel” < *breuđ-[6]  (cp. OHG brodi “fragile”) shifts its semantics to ME bread “bread,” supplanting earlier OE hlāf “bread.’ If we assume that PRom. *turta was loaned from a Celtic form from the British isles or Gaul and spread across the Romance dialect continuum, it must have reached East-Romance and Greek before the restructuring of the vowel system. We must remember that the Roman Empire in the heydays of the late principate and early dominate still constituted a cosmopolitan world in which goods, people and cultures quickly spread across the breadth of the Roman world.

                The variant Rom. *tarta can be found in ModFr. tarte “cake” and  ModSp. tarta “cake” alongside forms that must continue Rom. *tartara “cake” (ModIt. tartara “almond cake”, Comask. tártara “cake from milk, eggs and sugar” see REW 8590). Rom. *tartara looks like the feminine form of *tartaro meaning “wine stone” as found in ModSp. tártaro “id.” and ModFr. tairtre “id.” but its connection to the word “cake” is semantically very difficult. The connection between Rom. *tarta and *torta however is obvious. Nevertheless the vacillation in root vocalism prevents us from equating the etyma. To my mind we should take contamination with an etymon of Greek origin into account. The Greek word in question would then be the above mentioned Gk. ἄρτος “bread” which made its way to Western Romance as evidenced by Old Spanish artal “especie de empanada” and Basque arto “bread” (see Beekes 2009: 143). In combination with a preceding demonstrative resegmentation of the lexeme may have occurred, i.e. Rom. *est’ arto → *tarto (cp. OFr. icorne “unicorn” → l’icorne → ModFr. licorne, see Alkire 2010: 304-05) . This word would have easily taken its place alongside Rom. *torta “loaf of bread.”

Peter Alexander Kerkhof


Alkire, Ti et Carol Rosen, Romance languages; a historical introduction (2010).

Adams, J. N., The regional diversification of Latin 200 BC – 600 AD (Cambridge 2007).

Bevan, Gareth A. et P. J. Donovan, Geiriadur Prifysgol Cymru LVII (Cardiff 2001).

Bondarenko, Maxim Fomin Grigory, electronic Dictionary of the Irish Language (2002).

Bloch, Oscar, Dictionnaire Étymologique de la langue Française (Paris 1932).

Matthews, John Frederick, The Journey of Theophanes: Travel, Business, and Daily Life in the Roman East (Yale 1999).

Meyer-Lübke, Romanisches Etymologisches Wörterbuch, Sammlung romanischer  Elementar und Handbücher III (Heidelberg 1911).

Morris Jones, J, a Welsh Grammar: Historical and Comparative (Oxford 1913).

[1] In ModDu. we also find ModDu. toert which must derive from OFr. tort, tourte.

[2] For the common etymology, see which comprises the lemmata of the main etymological dictionaries of Dutch or which contains information from the main etymological dictionaries of English.

[3] An edition of the tablet may be found via this link:

[4] A compressed version of the dictionary can be found on:

[6] Verner-variant OE breað “brittle” < *breuþ-)

In Voce Gallica

A completely Gaulish sentence in the Vita Symphoriani Augustodunensis


The Gaulish language is mainly attested in epigraphical inscriptions from the 3d c. BC to the 2nd CE written in the Greek and Latin alphabet. During the principate (27 BC – 284 AD) Gaul was thoroughly Romanized and the Gaulish language lost ground to the more prestigous Latin/Romance language. Nevertheless Romance and Gaulish may have been spoken next to eachother well into the 5th c. and assuming the examplar of the Gaulish-Latin glossary known as Endlicher’s Glossary (Öst. Nationalbibliothek, MS 89) was composed in the 6th c.[1] we must conclude that even then some knowledge of the dying Gaulish language persevered.

Gaulish Chamalieres inscription

5th c. Saintslife

In the early 20th c. Wilhelm Meyer identified a Gaulish sentence in the oldest manuscrits of the 5th c. vita Symphoriani augustodunensis which relates the martyrdom of Symphorianus of Autun in approxamitely 180 CE. When Symphorianus is led to his place of execution his mother admonishes him from the city wall. Meyer reconstructs the passage as follows (Meyer 1901: 162).

uenerabilis mater sua de muro sedula et nota illum uoce Gallica monuit dicens: ‘nate, nate Synforiane, †mentobeto to diuo†’

‘his venerable mother admonished him from the wall eagerly and notable to all (?), saying in the Gaulish speech: “Son, son, Symphorianus, think of your God!”

corruption of the gloss

CLM 4585 f. 13

The passage has been corrupted in the manuscript tradition with most medieval manuscripts reading “nate, nate, synphoriane, in mente habe deum tuum.” They leave out the linguistic specification in uoce gallica and any words that are not understandable as Latin . Meyer’s reconstruction of the prototext was based on three manuscripts which represent an early redaction of the passion. The two CLM manuscripts are dependent on eachother, CLM 22243 being three centuries younger than CLM 4585[2].

                Turin D.V. 3          (9th c.)     nati nati synforiani mentem obeto dotiuo

                CLM 4585 , f. 13    (9th c.)     nate nate synforiane memento betoto diuo hoc e memorare di tui

                CLM 22243, f. 27   (12th c.)   nate nate symphoriane memento betoto diuo hoc e memorare di tui

Interesting in CLM 4585 is the Latin translation following the sentence which was probably part of the original 5th c. text. It indicates that when the passion was codified the Gaulish was no longer readily understable to all, although the sentence in question apparently had been part of the oral tradition surrounding the saint. Considering Jerome’s comment from around 386 that Gaulish was spoken in the area of Trier it is very plausible that Gaulish survived into the 5th c. and that also in the area of Autun the Gaulish language had persevered in order for the sentence to have been entered into the passion. This would mean that the sentence reflects a late 5th c. Gaulish.

The Gaulish sentence analyzed

The Gaulish was analyzed and interpreted by Rudolf Thurneysen in the Zeitschrift fùr Celtische philologie 14. The meaning of the Gaulish words can be listed as follows:

 Nate, nate, mentobe(to) to deuo.

Son, son, remember your god


nate < Gaul. gnate. The word would be in the vocative singular. We may assume that initial gn- just as medial –gn– has changed into palatalized Romance *ɲ– which was spelled as n– in initial position. It is not very likely that  nate refers to the Latin word (g)natus. The Latin word with the meaning “son” is already quite rare in Classical texts (probably because of competition with filius) and almost absent in Late Antique and Medieval Latin. The use of such an archaic word in a colloquial phrase would be highly unlikely. The Gaulish word gnatos however, seems, judging on its occurence in the inscriptions, to be the general word for “son”. 


Since the sentence is glossed in MS München 2223 with hoc est memorare dei tui the word must mean “remember” or “keep in mind.”


to is best explained as continuing PCelt. *towe possessive pronoun (cp. Old Irish do “your”), i.e. PCelt. *towe > Gaulish to. 


Gaulish dewos is well attested in inscriptions (e.g. PN dewognata and theVercelli inscription tewoxtonion “from gods and men”) . The noun would here stand in the accusative as well, i.e. Gaulish *dewom > dewo. The final nasal would have been lost in parallel to the loss of final nasals in Romance. The spelling diwo would show orthographical Romance confusion of short i and long e which had coalesced in PRom. *e. Once again, it is highly unlikely that we are dealing with the Latin word divus here, since the Latin word is archaic and hardly attested in late and medieval Latin texts.


In his analysis Thurneysen confirmed Meyer’s suspicion that the sentence was Gaulish but disagreed with Meyer about the word mentobeto. Meyer argued that the word may be a Celtic derivation to the same root as Latin memini “to remember.” This, however, seems unlikely since the verbal reflex of the root is only reflected in Old Irish muinithir “to intend” < PIE *mani̯e- (see De Vaan 2008: 371)[3]. Thurneysen therefore argued that mentobeto more likely represents a Vulgar Latin verb (Proto-Rom. *mɛntaƀerɛ, see REW 5507) derived from  Latin in mente habere (cf. OFr. mentevoir “to mention, to cite”, OProv. mentaure “id.”). This would mean that the sentence would be a mixture of Gaulish and Latin, a conclusion that is repeated by Adams in his book The regional diversification of Latin (Adams 2007).

Thurneysen argued that the form *mɛntaƀeto which would lie behind the spelling mentobeto would continue a Latin future imperative in –to.  Unfortunately the future imperative cannot be reconstructed for Proto-Romance as it is not continued in any of the Romance dialects, so the survival of the form in this late antique hagiography would be surprising[4]. It is highly unlikely that future imperatives still existed in 5th c. We should also note that the translating gloss memorare dei tui does not use the future imperative, which would be unexpected when the meaning of the Gaulish verb would exactly be such a future imperative[5].

Furthermore this would mean that we would have a Gaulish sentence, complete with a Gaulish direct object and possessive adjective, but with a Gallo-Romance inflected verb. This seems like an unlikely scenario. This indicates another explanation of the Gaulish inflectional form seems warranted. Alternatively we may argue that the word reflects a Gaulish inflectional form of a Gaulish verb borrowed from Romance *mɛntaƀerɛ < Latin mente habēre “to have in mind”. The –o- from attested mento- might reflect earlier *-a- that has been coloured by the following –ƀ- or we may speculate that the stem was adopted from Gallo-Romance *mentaƀ- as Gaulish *mentawƀwith the –aw- sequence monophtongizing to –o-. In either scenario I will propose to read the word as mentobe with the final to as a corrupted repetition of the following to.

I argue that mentobeto must be read as mentobe and represents a Gaulish imperative singular, namely Late Gaulish mentaƀe “remember!” < Early Gaulish mentaƀi (cf. Gaulish moni “come” < PIE *monH-e, Gaulish gabi “take” < QIE *ghHbh-e). This possibility assumes that the Gaulish inflectional form goes back to Romance *mɛntaƀerɛ. Naturally this scenario  also allows for the possibility that the word is Gallo-Romance after all, since a Romance imperative singular would also be *mentaƀe.


To conclude; it seems plausible that the un-latinate admonition in the 5th c. vita symphoriani augustodunensis  reflects a sentence in colloquial Late Gaulish. All the words are understandable as Gaulish, provided that we recognize the fact that mentobeto probably reflects a verb loaned from Gallo-Romance *mɛntaƀerɛ. The word mentobeto should be parsed as mentobe=to, with a repetitive to following the verbal form; this corruption is to due to the faulty tradition of the sentence. When the words were no longer understood scribes tried to make Latin out of them, interpreting the words as containing Latin memento. Actually the inflectional form mentobe would reflect Late Gaulish mentaƀe which is to be understood as an imperative singular, with the Gallo-Romance –a- darkened to –o- under influence of the following labial. This would connect neatly with the following Latin translation which translates the verb with Classical Latin memorare dei tui. Formally it is also possible that this mentobe is Romance after all,  namely reflecting Romance *mentaƀe, but it seems preferable to go for a Gaulish interpretation since all the other words do seem to reflect Gaulish vocabulary and grammar.

 P A Kerkhof


Adams, J. N., The regional diversification of Latin 200 BC – 600 AD (Cambridge 2007).

Eska, Joseph, “Continental Celtic,” in: The ancient languages of Europe, Roger D. Woodward ed. (Cambridge 2008) 165-188.

Matasovic, Ranko “The origin of the Old Irish f-future” in: evidence and counter-evidence, festschrift for Frederik Kortlandt (2008) 361-366.

Meyer, Wilhelm “Das älteste keltische Sprachdenkmal,” in: Fragmenta Burana, Festschrift der köningliche Gesellschaft der Wissenschaften in Göttingen (Göttingen) 1901 161-163.

Meyer-Lübke, Romanisches Etymologisches Wörterbuch, Sammlung romanischer  Elementar und Handbücher III (Heidelberg 1911).

Thurneysen, R. “Irisches und Gallisches,” in: Zeitschrift für Celtische philologie 14 (1923) 1-17.

Vaan, Michiel, Etymological Dictionary of Latin and the other Italic languages (2008).

[1] This seems reasonable considering caio (Gaulish *kagi̯o) in Endlicher’s Glossary is glossed with breialo sive bigardio, the latter word being a Frankish loanword which plausibly entered Gallo-Romance in the 6th c. only

[2] The CLM manuscripts are digitalized and can be found on

[3] We do find an e-grade to *men- in Old Irish toimtiu < PCelt. *to-menti̯on, so possibly a denominal formation analogous to Latin mentiō also existed in Gaulish. This however would probably have yielded a Gaulish verbal stem *mentii̯o- which does not fit with the attestation mentobeto.

[4] The early disappearance of the future imperative in to was probably provoked by the confusion with the past participle in –to < Lat. –tus.

[5] A Gaulish future is also unlikely since the Old Irish f-future cannot go back to an older *­-b- and is therefore not identical to the Latin b-future. The future suffix can be identified as Proto-Celtic *-iswā- (see Matasovic 2008: 361; contra Kortlandt).