Over at the Log, Mark Liberman quotes a spiky and suggestive short story by D. Barthelme, “They Called for More Structure.” He does so in the context of an analogy to syntax in machine translation that is neither comprehensible nor interesting to me; what is interesting to me is the punchline of the story, where the workers “saw the new city spread out beneath us, in the shape of the word FASTIGIUM. Not the name of the city, they told us, simply a set of letters selected for the elegance of the script.” Huh, I thought, I studied Latin and I don’t remember the word fastigium. So I looked it up in my handy paperback dictionary and found this farrago of senses: “gable; pediment; roof, ceiling; slope; height, elevation, top, edge; depth, depression; finish, completion; rank, dignity; main point, heading, highlight (of story, etc.).” This is how you can tell if you really like a language: if I were dealing with a Greek or Russian word, I would be pleased at the complex semantics, but since it’s Latin I just groan and think “Does any language really need a word like that?” (The first i is long, by the way, and apparently there is no clear etymology.)
It’s been borrowed into English (per the OED in an entry first published in 1895 and not updated since) in the senses “The apex or summit; spec. in Archit. the ridge of a house,” “The gable end (of a roof); a pediment,” and “The acme or highest state of intensity (of a disease)”; the last is the only sense given in the AHD.
Aww, Latin is my favorite language. Anyway, that’s right out of Aeneid 1.
No knock on Latin — I enjoy it and am glad I studied it! It’s just that I want it to stay nice and simple so I can understand it without too much difficulty. Omnia Gallia est divisa, that’s the kind of thing I like.
All Gaul is quartered into three halves, eh?
Actually, when I started second-year Latin my teacher was at pains to explain that omnis Gallia actually meant not “all (of) Gaul” but something like “Greater Gaul”, divided into Aquitaine (he taught French, too), Belgium, and Gaul proper.
Or to put it gastronomically: “All Gaul is divided into three parts: the part that cooks with lard and goose fat, the part that cooks with olive oil, and the part that cooks with butter.” —David Chessler
But to return to the farrago, if we think of it as meaning ‘high/low point’ (as altus means either high or deep in Latin, those being the same from different points of view, what might be glossed as ‘vertically extended’), I think all of the senses can be seen as straightforward metaphorical extensions. I was puzzled for a moment by ‘finish, completion’, but that must be because the ridge-pole goes on last.
Well analyzed!
In my understanding, fastigium is clearly derived from fastus, as in dies fastus.
All these words are closely linked with unique Latin concept of fas – divine law.
Latin religion is not really understood all that well.
We just don’t know what they really did in their temples (on gable end of a roof)
My trusty school Cassell’s gives fastigium as ‘the gable end, pediment of a roof’
with metaphorical senses:
1. a slope, either up or down
2. of measurements, looking up, height; looking down, depth
3. high rank, dignity
4. principal point in a subject
That Language Log entry was the first one that actually made me feel stupid. There have been several that were beyond my technical understanding, but those were clearly just things I would need to learn more about before I could get them. That one was just opaque to me. Or was that the point?
Fascinating! When talking about calculus (the modern one, not a Latin word) we do not relate extremum to large slope because it’s either goes with 0 slope (rounded top or bottom), or no slope at all (corner or something more difficult) or the end of the interval. But apparently Romans had some other imagery in mind, like pagoda, where apex is also a point of the largest slope. The latter situation in terms of modern calculus goes into “no slope” file.
On the etymology – FWIW, de Vaan (s.v. fa:sti:go, pp. 203-204) notes it as based on a Proto-Italic *farsti:g- / *farsti-ag-, ultimatly going back to PIE *bhrsti- “top, point”, which would make it cognate with English “bristle”. If I recall my Latin historical phonology correctly, the loss of /r/ would be irregular.
Loss of *-r- in a pre-Latin environment *-VrCC- such as in *-rst- would be regular if from earlier *farsti-. Compare Latin poscō “I ask for, call for, demand”, from earlier *porskō, ultimately Proto-Indo-European *pr̥(ḱ)-sḱe-, the *-sḱe/o- present of *preḱ-, “to ask”. My favorite example of the loss of *-r- in this kind of environment is the etymology usually proposed for Latin cēna “meal”, which goes along something like the following lines: cēna, from *kesna, from *kerssna (compare Oscan KERSNÚ and Umbrian çesna “meal, dinner”), from *kertsnā, from a PIE *kert-s-nh₂, “thing cut off, portion”, a derivative of the root *(s)kert- “to cut”.
Very nice indeed!
@ Patrick Taylor: Thank you for reminding me of posco: and ce:na; so yes, there are more example ofa loss of /r/ in similar environments. I clearly need to brush up on the PIE to Latin sound laws.
In the form “fastigiate,” this word is not uncommon in horticultural contexts. It refers to a tree or shrub that grows in a narrow, columnar, even spiky shape. “Vertically extended,” in fact.
The OED entry for fastigium was revised in 2021 and now has these senses:
(I like the vague specificity of “Now somewhat rare.”) And the etymology:
Oh, and the entry for fastigate ‘make (something) taper up towards a point’ (revised the same year) reads:
I have to say that last bit is ridiculously laconic, considering they’ve got all the space in the world online. How is it “probably < the same Indo-European base as brust” (‘bristle’)?
“Hurry up and pile them rocks, and make it fastigated!”
How is it “probably < the same Indo-European base as brust” (‘bristle’)?
While the OED could have given more details, that was discussed further up in the thread.
Mathematics seems to have for some slightly gratuitous reason adopted this word for a Johnson solid called the gyrobifastigium.
that was discussed further up in the thread.
I know, I wasn’t asking for information but expressing my irritation at their not going into that kind of detail.
Fastigium, fastigiate were included in the first edition of AHD (1969) among the descendants of PIE bhar- ‘Projection, bristle, point’ in the Appendix; the other cognates listed (in its erring-on-the-side-of-inclusion way) were: bass (fish with spiny dorsal fin), bristle, bur (prickly seed covering), brad, embroider, fastidious “from the notion of prickliness”, borscht originally made from “cow parsnip (from its sharp leaves)”.
The standalone edition of the Appendix by Calvert Watkins (1985, 2000) has all of these, except embroider; this is from French, and French broder is generally said to be from Germanic, but TLFI seems to be doubtful over whether this is from a verb meaning ‘prick (with needle)’ or from some other verb meaning ‘decorate, ornament’, or related to border?
But this root was another one that mysteriously vanished from the AHD from 1992 onward, even though Mallory/Adams and OED still think it’s valid (and Wiktionary: *bʰers-). So now their etymology of fastigium stops at Latin.
The only other one of that cognate list that has yet been revised in OED3 is fastidious (2021), and they reject the connection: they derive Latin fastīdium ‘disdain’ from fastus ‘pride, arrogance’ + taedium with haplology. This is not the fastus in SFReader’s comment above, which was second declension, but a fourth-declension noun, of unknown origin according to OED.
Best-case scenario with the OED entry is that it’s only a cross-reference and a fuller discussion will eventually appear under bristle, where from 1888 they have:
The starred protoforms will be deleted *spit*, but maybe they’ll add the Latin and Slavic cognates. Fingers crossed.
How is it “probably < the same Indo-European base as brust” (‘bristle’)?
There are more than 1200 instances of “the same Indo-European base as”. I think what’s happening is that the OED3, like Etymonline, is systematically removing all reconstructions, on the grounds that on the OED’s timescale they may have been re-reconstructed before the words are revisited, and sticking to actually recorded forms.
If you’d clicked on the link I provided, you would have seen that Language Hat is well aware of that, and that Patrick Taylor has argued that it’s a bad policy. The policy was stated in the Preface to the Third Edition, without giving a reason; there’s probably some supporting reasoning in the Philip Durkin article cited there, I’ll have to look it up, but I suspect it would be fair to summarize it as “they felt really burned by ‘no language has changed more than PIE in the last hundred years’ ”.
Why they have more confidence in Proto-Algonquian reconstructions than Proto-Indo-European is another mystery.
ETA: open access to Philip Durkin’s Root and Branch: Revising the etymological component of the Oxford English Dictionary (1999) (click the PDF link for the full paper).
Besides, you can provide a lot more information than “the same Indo-European base” without actually giving a reconstructed proto-form. It feels like passive-aggressive withholding.
Even if the reconstructions are “wrong”, they provide (some) convenient immediately visible information as to how the attested forms are related. If you suppress them, you are (for example) either asking your readers to believe, on mere authority, without any supporting explanation, your implausible assertion that (say) “ten” is cognate with Sanskrit “daśa”, or expecting them to work out Grimm’s Law etc for themselves on the spot. I don’t see the point of mentioning the Sanskrit form at all if you’re going to do that. With that attitude, the dictionary should just confine itself to tracing loanwords when it comes to etymology.
This decision may reflect an understanding of reconstructions which is too slanted to interpreting them as directly representing a real spoken protolanguage rather than as convenient formulae for expressing cognacy. This is particularly silly given the highly abstract nature of current PIE reconstructions, which look nothing at all like pronounceable words in a real human language, really.
It can feel like passive-aggressive withholding now, but I think it’s inevitable when they’re publishing piecemeal updates. Nobody expects them to repeat the full discussion under each and every cognate, right? I personally don’t think they did anything wrong in giving only a cross-reference under fastigium; you can’t get a more complete story yet since this update wasn’t synched with those for bristle and brust, but those will come out sooner or later.
Whether to present starred reconstructions is a separate question.
Nobody expects them to repeat the full discussion under each and every cognate
with them working in a digital format in the age of cut-and-paste, that doesn’t actually seem like a particularly onerous expectation, or even a particularly odd one.
This decision may reflect an understanding of reconstructions which is too slanted to interpreting them as directly representing a real spoken protolanguage rather than as convenient formulae for expressing cognacy.
Is this a widespread understanding ? That’s the way they come across to me. Of course I myself don’t indulge.
This is particularly silly given the highly abstract nature of current PIE reconstructions, which look nothing at all like pronounceable words in a real human language, really.
That final “, really” sounds like a cautiously recommended Reality Check. Given the way many people here go into great detail, at even greater length:
# Loss of *-r- in a pre-Latin environment *-VrCC- such as in *-rst- would be regular if from earlier *farsti-. Compare Latin poscō “I ask for, call for, demand”, from earlier *porskō, ultimately Proto-Indo-European *pr̥(ḱ)-sḱe-, the *-sḱe/o- present of *preḱ-, “to ask” …# [from upthread]
one could be forgiven for feeling that they do take reconstructions seriously. Like bridge experts discussing bidding systems, or military procurers discussing bedding systems.
To paint these activities as “merely analytical” doesn’t match the zeal with which they are pursued, and could be resented. Thus your caution, I imagine.
in a digital format in the age of cut-and-paste
Anybody who’s tried to maintain a large piece of software, or a very large reference work such as the OED, knows that copy-paste is risky. What if you want to make a change in *one* of the copies? How are you going to make sure it’s propagated to all the others? It’s a very common problem.
It was already the digital age of copy-paste in 1999 when Durkin was writing; he devotes a lot of space to the policy of using cross-reference rather than repetition, and I think he was right on that one.
Yeah, copy-and-paste isn’t really enough to keep things up to date and harmonized. The “right” way to do it is to have all the data drawn from a relational database. For example, an online dictionary would have an entry for each modern lexeme, which refers to its direct etymological antecedent(s); each antecedent has its own entry; and so on. That way, if you change the etymology of the Latin form from which a word is derived, that propagates to every other etymology that is based upon it.
The problem is that implementing this gets complicated quite quickly. Moreover, the complexity makes it especially difficult to add functionalities that were not part of the original plan.
What if you want to make a change in *one* of the copies? How are you going to make sure it’s propagated to all the others? It’s a very common problem.
Another common problem that I have to deal with, is wanting to make a change in one of the copies *without* having it propagate to the others. “eclipse” is an IDE (integrated development environment), a complex program with hundreds of variable settings. The eclipse people developed a system for automatically propagating changes in one eclipse “workspace” to all the others you have (“Intelligent Software”).
I don’t want that, and fight to keep it from happening, but it’s not easy.
The classic victim of automated propagation is Mr. Mickey Mouse in “The Sorcerer’s Apprentice”. Part 1, Part 3.
Developers call this “continuous integration”, which is designed to proceed without prodding. Once it gets going, they don’t know how to stop it at a desired point (“debugging”). They end up having to wait for 5 minutes after coding for 1 minute. That seems to be OK with them, they have a latte and brush up on their buzzwords in the meantime.
…but not in the last 40 or so.
Probably for the same reason: there aren’t as many as of PIE – more like one or two, AFAIK.
IEists do have a highly annoying habit of using eclectic mixes of phonetic, phonemic and morphophonemic transcriptions, sometimes in the same reconstructed word. They’re individually different, too.
This impression is greatly exaggerated by 1) morphophonemic reconstructions of stems with hyphens and without endings or sandhi; 2) not writing the non-phonemic schwa which can be reconstructed in a lot of positional detail (…though I disagree with one of the rules proposed in that thesis because of this later work…); 3) continued insistence on not postulating any sound values for the “laryngeals”, which made some sense 40 years ago but not now.
The consonant clusters PIE actually had were quite a bit less formidable than those of Polish, on average.
@Brett: I think an RDB would be overkill in principle, from the very start. You don’t need transaction control, because there is hardly any write contention. Writes are infrequent. They can be planned and coordinated in advance. The database is mostly read-only. This is a job for NoSQL.
Unlike the case for an airline booking system, where every minute thousands of seats are reserved and cancelled. There is no possibility of planning and coordinating in advance. That’s why transaction control is needed.
I doubt that at the OED there are thousands of eager-beaver contributors who every minute introduce and delete *spit* proto-forms in the same entries.
Is this a widespread understanding ? That’s the way they come across to me
We were not long ago actually discussing this question of what, exactly, people think they are reconstructing in comparative work. My Google-fu is not currently up to locating the discussion though.
My own feeling is that PIE was surely a real language, but our reconstructions are theoretical constructs, and thus not real at all, so it’s a mistake to get too obsessive and puritanical about whether they are “correct”: all they can aspire to is being useful, and they just need to be good enough for the purpose in hand. It is not the purpose of the OED to reconstruct PIE, or to demonstrate that Indo-European languages are all genetically related, or to establish the subgroups within Indo-European rigorously.
Still, it’s not an all-or-nothing dichotomy: grammatical analyses of real modern languages are also theoretical abstractions, for all the Chomskyite striving to reify them, so it’s a question not of abstract versus real but of degrees of abstraction.
And the assumption that a protolanguage actually was a historical real spoken language at one time is also extremely useful as a heuristic in theory-building: it should stop your reconstructions from ending up like something that can’t be interpreted as reflecting any known sort of human language at all. (Someone should give some of the reconstructors of proto-Central Chadic a good talking-to on this point.)
It should also instill a proper scientific humility (without which, progress is impossible): nobody has ever devised a protolanguage by comparative methods that covers all that constitutes a complete grammar. Even the most detailed reconstructions are manifestly incomplete things compared with a competent study of any modern language. Inevitably.
The take-home message is thus: “hoe in humility”. That’s a tall order, but as it happens I agree with you.
Everyone admires the principles of moderation set out in Sunday’s sermon, but during the week you couldn’t tell that was the case. They think Calvinism has something to do with Calvin Klein.
They confuse Calvinism with Calvin Klein
It’s easily done. Calvinism is just so modern and chic. You can understand people falling for the sheer glamour.
David E: asking your readers to believe, on mere authority … or expecting them to work out Grimm’s Law etc for themselves
The OED doesn’t expect readers to believe on their authority, they expect readers to already know Grimm’s law and where to go to look up sound changes between PIE and Latin. Page space may not be limited, but editors’ time and budget are, and they have thousands of other etymologies due every quarter. Are they responsible for explaining all relevant sound laws in every etymology? Durkin says no — a dictionary is not a text on historical linguistics — and I find it hard to disagree, though I think a citation to de Vaan in the fastigium etymology could have been practical for them, and helpful for us.
(Hans [February 24, 2015]: de Vaan on the page you cite *does* say the loss of r was regular by “Schrijver’s rule”, which I think is what Patrick Taylor described in the next comment?)
AHD, of course, is the dictionary for readers who can’t yet work out for themselves how ten is cognate with Sanskrit dáśa, and want to learn.
You might have more ground for accusing Merriam-Webster, Random House, NOAD, and the New World of asking their readers to “believe, on mere authority”, since they do include IE cognates, and sometimes even reconstructions, without an accompanying tutorial on the comparative method. More charitably, I think what those publishers actually expect is that their readers may be intrigued enough by the romance of ancient cognates to go find a book about historical linguistics.
You raise a good question on how useful the OED’s piling-up of supporting evidence (Sanskrit etc.) really is to readers when they exclude all theoretical interpretation (sound laws, protoforms).
On TikTok the Calvinists are always dissing the Armanians
The Armenians are not so bad. A bit too keen on speaking in tongues, but other than that…
What if you want to make a change in *one* of the copies? How are you going to make sure it’s propagated to all the others? It’s a very common problem.
i could take this argument seriously in a lot of contexts, but while talking about OED revisions, where the anticipated timeframe for just about anything is likely to be expressed in decades? nope. any kind of comprehensive harmonization across entries is already pure fiction.
rozele, do you spend much time studying the OED’s updates every quarter? I do, and the timeframe is not decades for “just about anything”: I’ve gotten corrections made within two months (though four to six is more typical). Harmonization across entries is not fiction, I’ve gotten them to fix inconsistencies in book titles and dates. If they had written a full description of the etymology of fastidium back to PIE with all cognates listed, and then had some modifications later in the full revision of bristle, it would definitely not be acceptable by their present editorial standards to leave the inconsistency dangling. Each quarter well over 100,000 entries get some sort of edit (about 90% are pretty minor tweaks). It isn’t 1972 anymore.
@ktschwarz: i’ll gladly take the correction!