Bracchium to Brezel to pretzel

« previous post | next post »

I'm in Frankfurt for a week, and a stroll through the Weihnachtsmarkt last night with Caroline Féry and Ede Zimmermann reminded me of something I've wondered about for a long time: Why was German Brezel borrowed into English with an initial 'p'?

So I looked it up. The OED's etymology for pretzel explains everything:

A borrowing from German. Etymons: German Bretzel.
< German Bretzel kind of bread roll, made from a thin length of dough twisted into a knot and coated with brine before baking (now usually Brezel ; Old High German as brēzila , Middle High German brēzel , prēzel , prēzile ) < post-classical Latin bracellus kind of cake or biscuit (12th cent.), shortened < an unattested post-classical Latin form *brachiatellus (compare post-classical Latin bracidelli (plural) bakery items (in an undated glossary)) < classical Latin brachiātus , bracchiātus brachiate adj. (compare post-classical Latin braciatus (noun) kind of cake eaten on monastic holidays (11th cent.)) + -ellus  suffix; so called on account of the resemblance to folded arms. Compare Italian bracciello a kind of cake, simnel, or biscuit (1598 in Florio).

Compare ( < post-classical Latin *brachiatellus) Old High German brēzitella, Old Occitan bressadel, brassadel kind of ring-shaped cake (1480; Occitan braçadèl type of cake made with eggs, cake in the shape of a braid), Italian bracciatello kind of ring-shaped cake (second half of the 15th cent., also as bracciatella).

The English form with initial p- probably represents a perception of the unaspirated pronunciation of b- in regional German (south.).

The idea that the 'p' was a misperception of an unaspirated /b/ had occurred to us. But the fact that Brezel (and therefore pretzel) comes from the Latin word for "arm" is an unexpected pleasure.



  1. Bob Ladd said,

    December 14, 2015 @ 5:17 am

    There are several words in German that vary in the voicing of initial stops, such as gucken / kucken 'look'. There's also the hard-to-translate word Pauschale, which means a variety of things from 'lump sum' to 'per diem', and which is etymologically related to the phrase in Bausch und Bogen, which the Collins bilingual dictionary translates as 'lock, stock and barrel'. So it's not just between English and German that such misperceptions occur, but within German as well.

    German phoneticians have been saying for ages that "voicing" is not really the basis of the distinctions involved here, but rather something like "fortis/lenis". In Swiss German it's well-established that these contrasts are based phonetically on closure duration and force of articulation, not closure voicing or post-release aspiration, but similar deviations from mainstream IPA expectations are found in other German varieties too, and must give rise to cross-dialect confusions from time to time.

  2. Guy said,

    December 14, 2015 @ 5:20 am

    This reminds me of my surprise when I first learned that the English word "plant" comes from the Latin word for the sole of a foot, though the semantic relation becomes apparent when you think of expressions like "planted her feet firmly in place on the ground".

    I imagine that body-part words are disproportionately prolific in etymologies.

  3. Phillip Minden said,

    December 14, 2015 @ 5:33 am

    I suppose you mean voiceless (but lax) Upper German b.

  4. Matt_M said,

    December 14, 2015 @ 5:56 am

    @Guy: the Online Etymology Dictionary says that Old English plante derives "from Latin planta "sprout, shoot, cutting" … perhaps from *plantare "to drive in with the feet, push into the ground with the feet," from planta "sole of the foot".

    So it seems that that etymology is at least provisional or disputed. There are other interesting descendants of the Latin word planta, though: the Welsh word plant, meaning "children" and the Irish word clann "children, offspring, clan", which follows a typical old Irish pattern of borrowing Latin /p/ as /k/.

  5. Vilinthril said,

    December 14, 2015 @ 6:09 am

    @Bob: “gucken” vs. “kucken” is a different cookie, though. In Southern German/Bavarian (incl. Austrian) dialects, there is no difference between word-initial b/d and p/t – for instance, the dialectal pronunciations of “durt(e)n” (= dort = there) and “Torten” (= cake(s)) is identical, as are “Pass” (= (mountain) pass) and “Bass” (= the voice type). However, “Garten” (= garden) and “Karten” (= cards) are very much different in their pronunciations.

  6. Enka said,

    December 14, 2015 @ 7:39 am

    This reminds me of a German sketch that makes fun of the regional dialect from Saxony:

    It plays on the difference (or the percieved lack of it when pronounced by a dialect speaker) between:
    "Parkettboden" (parquet floor) and
    "Baguetteboden" (a non-existing which would mean baguette floor)

    Of course this "cross-dialect confusion" would not occur in real life, but I can confirm similar misperceptions when listening to speakers from Saxony.

  7. Rachel B said,

    December 14, 2015 @ 8:10 am

    Some nice cross-cultural comparisons can be seen here: a "pretzel" is a representation of folded arms, and a "bagel" is, ultimately, something that's bent (OHG bouc- for ring, bracelet, which goes back to the Proto-Germanic stem *bug- "to bend, to bow").

  8. Christian Saunders said,

    December 14, 2015 @ 8:26 am

    In Spain they call them "bretzel"! At first I thought it was a typo, but then I started seeing it everywhere.

  9. Peter Erwin said,

    December 14, 2015 @ 8:39 am

    I was amused to note that some Munich-based colleagues, working on understanding the orbits of stars in our galaxy, decided they had identified an important new class of orbits, which — since "pretzel" was already taken for a different class of orbits — they decided to call "brezel"…

    "Peanuts, brezels and bananas: food for thought on the orbital structure of the Galactic bulge"

    ("Bananas" are yet another class of stellar orbits; there are also "anti-banana" orbits…)

  10. Coby Lubliner said,

    December 14, 2015 @ 10:20 am

    The Pennsylvania Dutch Wikipedia page reads "En Bretzel iss en Pretzel." Pretzels seem to have been introduced to America by the Pennsylvania Dutch.

    What I find interesting is that what is generally known as a pretzel was called "beygele" among Jews in Poland, and this usage seems to persist in Israel, since the Hebrew Wikipedia page corresponding to "Pretzel" is headed בייגלה. I was surprised to find out, on coming to the United States, that "bagel" refers more or less to what we had called (in Polish) obwarzanek. Vendors of obwarzanki in Cracow have also bajgle written on their carts, perhaps as a concession to tourists looking for bagels.

  11. Pflaumbaum said,

    December 14, 2015 @ 11:13 am

    I've often wondered whether the centrality of thee fortis/lenis distinction over voicing in German and other Germanic languages has any bearing on Grimm's and Verner's laws, which are always formulated in terms of voicing.

  12. Rod Johnson said,

    December 14, 2015 @ 1:10 pm

    Yeah, I was (am) confused by "unaspirated /b/." What exactly is the phonetic realization of the fortis/lenis contrast here?

  13. Lazar said,

    December 14, 2015 @ 1:42 pm

    I read recently that Donald Trump's family name was Drumpf; in that case, the change to t may have been influenced by the pre-existence of trump as an English word.

  14. Pflaumbaum said,

    December 14, 2015 @ 3:10 pm

    Yes he took his time joining the /t/ party.


  15. Bob Ladd said,

    December 14, 2015 @ 3:25 pm

    @Rod Johnson: I think the short answer is that nobody knows, at least not in detail. What does seems certain is that simply talking about voicing and voice onset time (which is in effect what the taxonomy implicit in IPA transcription does) isn't enough to describe the various articulatory differences that manifest "voicing" (or fortis/lenis) distinctions.

    One clear piece of evidence for this claim comes from the fact that in English, "voiceless" stops exert a local raising effect on fundamental frequency following the release of the closure, while "voiced" stops don't. This is true of both voiceless stops in syllable-initial position (like pit) and in sC- clusters (like spit), even though in terms of voice onset time the /p/ in spit is like the /b/ in bit – which doesn't locally raise F0. Whatever causes the local rise in F0, it's not well correlated with voice onset time. (Source: a paper in JASA by Helen Hanson – 2009, I think.)

  16. Not a naive speaker said,

    December 14, 2015 @ 5:04 pm

    More from Grimms Wörterbuch:

    Variations are Bretze or Bretzel

    Some serious fact poisioning from the Schweizerisches Idiotikon

  17. David Marjanović said,

    December 14, 2015 @ 5:28 pm

    In Southern German/Bavarian (incl. Austrian) dialects, there is no difference between word-initial b/d and p/t –

    Depends on which dialects. I'm from Upper Austria and do differentiate these, although 1) they're all voiceless and unaspirated and 2) initial /p/ tends not to occur in words of native origin. This is carried over into Austrian Standard German.

    High German started out with three series of plosives:
    – long fortes
    – short fortes (of which /p/ and /k/ occur mainly in loans and in positions where length is neutralized, leaving /t/)
    – short lenes
    (Long lenes occur across morpheme boundaries: |dd| and |td| show up as /dː/, while |tt| and |dt| join /tː/.)

    That's a rather crowded system; three series of plosives that are neither aspirated nor voiced nor glottalized nor whatever the Korean "fortes" are is a bit much. We should expect it to be unstable. Indeed, various things happened to it in most of the German-speaking area:

    1) North of the Upper German area, the length distinction was lost, leaving at most two series.
    2) The sound system of Standard German is just barely Upper German (witness /p͡f/), but the actual sounds people use to fill in these abstract slots depend on the dialects that are or were spoken where they're from. This means that voice (though not very reliably, like in English) and aspiration (replaced by glottalization in the same conditions as in English) show up in overlapping areas from about the middle of the Central German area northwards; both are codified in German stage pronunciation, which is half Low German accent, half deliberately crafted artificial features that are easier to hear than the natural alternatives.
    3) In Switzerland (roughly speaking), the long and the short fortes merged as long lenes. Swiss German thus has two series of plosives, with a length distinction (even at the beginnings of words!) that continues the lenis/fortis distinction.
    4) In eastern Austria, all fortes became lenes, but nothing happened to the length distinction: the short fortes became short lenes, not long ones. Like in Switzerland, only a length distinction is left, but it doesn't line up with the Swiss one; it continues the old length distinction, except where lost due to 5).
    5) In the rest of Central Bavarian (including my dialect), at least /t/ and /tː/ became /d/ between vowels*; likewise, the rare /pː/ became /b/ in at least one word, and /kː/ became /g/ in at least one word. This is not carried over into Austrian Standard German. Then, /b/ between vowels merged into /v/. In short, all three series persist, but they don't have many opportunities to contrast. – In addition, /t/ became /d/ at the ends of words if directly preceded by a vowel; this is carried over into Austrian Standard German. (We don't have syllable-final fortition! The very word Auslautverhärtung has such a /d/, mwahah hah.)
    6) The dialects in and around Carinthia (southernmost Austria), which belong to South Bavarian, have reinterpreted their whole sound system in Slovene terms. Consonant length is lost across the board, and the lenes are voiced; I think there's even word-final fortition.
    7) Binnendeutsche Konsonantenschwächung, "Inderior German Gonsonant Weagening": in a broad area that includes the Saxon**/Thuringian and Swabian dialects, both length and the fortis/lenis distinction were lost on the phonemic level, leaving a single series of plosives. In Saxony at least, fortes and lenes (all short) persist as allophones: fortes occur syllable-finally and before /r/ and /l/ (and I think before /n/, which is a rare occurrence), lenes everywhere else.

    Phonetically, the /p t k/ of Standard Italian, the Slavic languages and Japanese are unaspirated fortes, as are the French /p t/ (French /k/ is borderline); the /p t k/ of Spanish, Hindi and Thai are voiceless lenes, and so are the Mandarin phonemes transcribed b d g in Pīnyīn but p t k in Wade/Giles and several other transcription systems. Navajo b d g are voiceless lenes, too.

    Aaaanyway, back to pretzels (Brezeln). The voiceless lenes go back all the way to the High German Consonant Shift. When Old High German began to be written (shortly after this shift), it was written by people who were literate in Latin and (at least in some cases) familiar with local Romance varieties, so they tended to use p t c/k for the lenis plosives. (Example: kepan, modern geben – "to give".) The resulting confusion was never completely sorted out, especially in the case of word-initial p, because /p/ isn't native there but soon entered the sound system in loans; p for etymological /b/ is common in Austrian placenames and surnames, and there are even first names with -pert instead of -bert.

    In short, the p in pretzel is most likely a confused spelling with ancient roots; alternatively, it could have been added as the word wandered through northern Germany or the Netherlands.

    * Everything in this paragraph happened after the vocalization of /l/ but before the vocalization of /r/.
    ** That's Upper Saxon (Sächsisch, Obersächsisch), spoken in Saxony just north(west) of the Czech Republic, not Low Saxon (Niedersächsisch), which belongs to Low German and is descended from Old Saxon. The name Saxony drifted southeast over a couple of centuries as feudal dynasties gained territory on one side and lost it on the other.

  18. David Marjanović said,

    December 14, 2015 @ 5:34 pm

    a "bagel" is, ultimately, something that's bent

    *Beugel, with regular Yiddish /ej/ for eu (long story), from beugen "bend (down)", "-flect" and a diminutive suffix.

  19. Jongseong Park said,

    December 14, 2015 @ 5:49 pm

    @Lazar, according to The Trumps: Three Generations That Built an Empire by Gwenda Blair, the Drumpf family changed the spelling to Trump during the Thirty Years' War when they were still based in Germany, in Kallstadt (p. 26). So the change has nothing to do with English. And there are other Trumps in Germany—diehard typophiles might have heard of the typeface Trump Mediaeval, designed by German typeface designer Georg Trump (1896-1985).

  20. Adrian Bailey said,

    December 14, 2015 @ 8:04 pm

    I was in Heidelberg and decided to look up an old friend. I phoned him up and he told me he'd moved to Blankstadt. I had trouble finding this place until I scoured the local area map and realised he'd said Plankstadt. Speaking to him later I'm sure his intial p and b were different, but to my English ear his p was closer to a b than to a p.

  21. Pflaumbaum said,

    December 14, 2015 @ 8:43 pm

    I remember as kid trying to read Asterix in French, and being baffled by the use of pon to represent bon as spoken by a Goth.

  22. Rod Johnson said,

    December 14, 2015 @ 9:57 pm

    Thanks, Bob and David. For some reason I overlooked Bob's initial comment at the top of the thread, where he answered the question I later asked. But the subsequent discussion was very interesting, thanks.

  23. Lazar said,

    December 14, 2015 @ 10:42 pm

    @Jongseong Park: Huh. When I checked Wikipedia a few months ago it said that Donald Trump's grandfather had been named Drumpf, but it's now been edited to say that that claim is erroneous.

  24. SlideSF said,

    December 15, 2015 @ 1:02 am

    I got my etymology from the back of the Snyder's of Hanover box when I was a kid:
    The Legend of the Pretzel-

    The Story of the Pretzel begins about 1500 years ago in a monastery; it is believed, in Southern France or Northern Italy. There, the monk whose task it was to bake bread found he always had scraps of dough left over when he'd shape the loaves. Being as imaginative as he was frugal, he formed these scraps into dough-ropes which he twisted into the now-familiar shape. Someone soon noticed how this shape resembled that of a child's arm folded in prayer, and since the demand had almost instantly outstripped the supply, our friend decided to called them "Pretiola" – "little rewards" in Latin – and saved them for the children who learned their prayers well.

    Apochryphal perhaps, but I have since seen it corroborated, for instance here

  25. flow said,

    December 15, 2015 @ 4:34 pm

    @David Marjanović—OMG now *this* is detail! The German dialectal situation is indeed quite puzzling in this respect. I guess the fortis/lenis distinction would make a good subject for a longer article, preferably one with many contrastive audio samples.

    I've always been puzzled why several authors of romanization systems for Chinese, Tibetan and Korean were so obsessed to insist on writing k/k' or k/kh on the grounds that the 'strength' contrasts in the plosive series—p/b, t/d, k/g—are (in all these languages?) realized with aspiration, not voicing. Among other things, this has given us "Peking" and "Taipeh" (with "Taipeh" being simplified from "T'aipeh"), which otherwise could have been "Beging" and "Taibeh" (the vowels and the later lenition of /gi-/ are another matter). When you have a language with a two-way 'strength' contrast in the plosives, it's just natural to use p/b to write that. Of course, when transcribing a language with three or more contrasting series or when comparing phonetic details across languages, the notation must be more complex.

    Incidentally the German Wikipedia article on 'fortis' ( has a few examples of German to Czech language transfer where South German b, d, g surface as p, t, k in Czech: piglovat from bügeln, pichle from Büchlein, herkot from Herrgott, purkrabí from Burggraf, tucet from Dutzend. Maybe the Czechs brought the Pretzel across the Channel?

  26. David Marjanović said,

    December 15, 2015 @ 4:46 pm

    I've always been puzzled why several authors of romanization systems for Chinese, Tibetan and Korean were so obsessed to insist on writing k/k' or k/kh on the grounds that the 'strength' contrasts in the plosive series—p/b, t/d, k/g—are (in all these languages?) realized with aspiration, not voicing.

    It makes perfect sense if you're French, for example. Voiceless lenes seriously confuse Russians and are routinely understood as fortes by French.


    …has the added factor that [g] isn't native there: it turned into [ɦ] long ago. Because /k/ has nothing to contrast with, it could turn into [g̊]; I've encountered the claim that it has, but I haven't heard enough Czech to confirm this.

  27. David Marjanović said,

    December 15, 2015 @ 4:49 pm

    The German dialectal situation is indeed quite puzzling in this respect.

    Hey, the whole point of the High German Consonant Shift was to get rid of voice and aspiration in all obstruents. :-)

  28. Lazar said,

    December 15, 2015 @ 9:43 pm

    @flow: Yeah, it puzzles me that people insisted for so long on avoiding b, d and g in Mandarin transliteration. The distinction between apostrophized and non-apostrophized letters in Wade-Giles was bound to be ignored by regular people (as, indeed, it was in postal system pinyin), resulting in a whole bunch of neutralizations on top of the customary omission of tones. "Kung", in popular transliteration, could correspond to any of eight different syllables in Mandarin.

  29. Jongseong Park said,

    December 16, 2015 @ 2:13 pm

    The Korean fortis consonants are voiceless unaspirated fortes. In word-initial position, the three can be distinguished by:

    Fortis – unaspirated / high burst energy / creakiness / higher pitch
    Aspirated – heavily aspirated / high burst energy / higher pitch
    Lenis – lightly aspirated / low burst energy / lower pitch

    The creaky effect on the following vowel (measured by spectral tilt, h1-h2) is what distinguishes Korean fortis consonants from similar voiceless fortes in other languages, but the difference can be overstated. In medial position, the secondary cues (relative burst energy, pitch, creakiness) are no longer needed, and you have a simpler, cross-linguistically common contrast:

    Fortis – unaspirated voiceless
    Aspirated – aspirated voiceless
    Lenis – voiced (unless next to a voiceless sound, in which case it neutralizes with fortis)

    Korean speakers will by and large hear not just fortis voiceless unaspirated sounds but what David Marjanović calls voiceless lenes—the /p t k/ of Spanish, Hindi and Thai, and the Mandarin phonemes transcribed b d g in Pīnyīn but p t k in Wade/Giles—as Korean fortis sounds (although the Mandarin phonemes with lower tone will also be heard as lenis sounds).

    The common practice of transcribing initial lenis sounds in Korean as /p t k/ is therefore very misleading, especially next to /pʰ tʰ kʰ/ for the aspirated series (for one thing, it seems to be a common mistake to assume that Korean lenis sounds are unaspirated, when they actually are always lightly aspirated initially). I believe that /p t k/ in Korean should be used for the fortis sounds, writing initial lenis sounds as /b̥ d̥ ɡ̊/.

  30. Davide said,

    December 16, 2015 @ 4:36 pm

    what I can say is that if you want to mock the Italian pronounciation of German people you substitute /b/ with /p/, i.e. "bambino" comes out as "pampino" because that's what we hear when they speak Italian.
    However isn't there the same problem of voiced/voiceless with the pron. of German "tag" (day) as /tak/ ?

  31. David Marjanović said,

    December 16, 2015 @ 7:03 pm

    That's very interesting about Korean, thank you!

    I can't seem to pronounce an aspirated voiceless lenis, though. Are they actually clusters with [h]? The initial cluster /gh/ [ɡ̊h] is very common in my dialect and distinguished from both /k/ and /g/; minimal triplet: Kern "kernel, core", gern (adverb for liking to do something), gehören "belong" (with syncope and unrounding).

    However isn't there the same problem of voiced/voiceless with the pron. of German "tag" (day) as /tak/ ?

    Yes and no… the g is voiceless for everyone, but in the north what's going on here* is that it's reliably voiceless only because it's at the end of a syllable, where /b d g/ are completely replaced by /p t k/. This is called "syllable-final fortition"** and generally absent from southern German accents.

    * Never mind the common pronunciation [tʰaχ], which applies the same process to Low German /ɣ/…
    ** Well, most often it's called "word-final devoicing", but that's doubly misleading.

  32. Jongseong Park said,

    December 17, 2015 @ 3:21 am

    @David Marjanović, I wouldn't say that the Korean lenis consonants are clusters with [h] the way that the aspirated series could be thought of as. When I say that they are lightly aspirated, I simply mean that they have a positive VOT. This does not mean that the aspirated portion is prominent enough as is the case with the aspirated series, which have indeed been analyzed as clusters with /h/, with allophonic variation depending on following vowels—[ç̞] before /i/, [ʍ] before /u/, [x̞] before /ɯ/, etc—just like /h/ on its own. I think of the lenis series at initial position as whispered versions of voiced consonants, which causes the delay in voicing.

  33. David Marjanović said,

    December 17, 2015 @ 4:44 am

    I see.

  34. flow said,

    December 17, 2015 @ 7:57 am

    @Lazar—I think I can top that. Just think of Wade-Giles chun, ch'un, chün, ch'ün—four basic syllables that often get contracted to a single chun, leading to no less than 16 distinct syllables that all get spelled the same. Of course, there's a similar problem with ü in Pinyin (where it is even more confusing because the orthography specifies the dots have to be written after n, l but not after q, j, y).

  35. January First-of-May said,

    December 21, 2015 @ 9:32 am

    @ Cody Lubliner, on pretzels:

    The traditional Yiddish song "Beygelayn" seems to equate the titular product with "bublichki" (or maybe the other way around? I know it uses both words for it), which is the same root as the Russian word "бублик" (which means, according to Wikipedia, the same as the Polish "obwarzanek").

    Generally, I'm not sure of the precise difference between sushki (сушки, literally "dry", basically a type of cracker), baranki (Russian cognate to "obwarzanek"), bubliki, bagels (in the American sense), and donuts (пончики) – it seems to be a continuum (roughly ordered by increasing softness), and the shape is definitely the same (a torus); but the word "pretzel" carries the mental image of "to twist in a pretzel", and the corresponding very different shape (the Russian for it is "крендель", though the Wikipedia articles are separate – the English cognate to that Russian word is apparently "kringle").

  36. Monscampus said,

    December 25, 2015 @ 9:59 am


    Your statement that Germans pronounce Italian words, e. g. bambino, as *p*am*p*ino really amazes me. Did you really hear any German speak Italian like that in real life?

    As a native I never heard this in any regional German dialect. On the contrary, certain dialect speakers are well-known for substituting a b for a p (when speaking German) as stated by previous commenters. Regarding the two different pronunciatons of Tag, that's a different matter. They are both non-dialectal standard German, for the most part on different sides of the so-called white saugage equator. /sprachwissenschaft/ada/runde_1/f15a-b/

  37. David Marjanović said,

    December 25, 2015 @ 7:06 pm

    Your statement that Germans pronounce Italian words, e. g. bambino, as *p*am*p*ino really amazes me. Did you really hear any German speak Italian like that in real life?

    You'll hear plenty forgetting to voice the [b]; and I'm sure lots of Italians would interpret [b̥] as /p/ at least half of the time.

    Voiced [b] is so unthinkable in most of southern German that German-speaking sheep make määäh. Personally, I actually hear [m] when I hear sheep.

  38. Monscampus said,

    December 25, 2015 @ 8:56 pm

    @David Marjanović

    Will I in future? Because I never did in the past. It must be my fault for hardly ever visiting Bavaria, but I heard lots of Germans speaking Italian including myself (or maybe me). I don't see why Northern German sheep should adopt a southern accent, though. It's määäh all over Germany. The bah-sounds uttered by sheep in the Anglosphere indicate that the animals suffer from a bad cold despite of all their wool.

RSS feed for comments on this post