Bill Walderman asked
Is it possible that the "rule" requiring the placement of "however" as the second element of a sentence (or at least not the first element) … originated as an attempt to impose Latin and Greek syntax or word order on English? In Latin, there are a number of particles such as "autem" that can't be placed at the beginning of a clause and usually appear as the second element. And Ancient Greek is extremely exuberant in particles that must be placed as the second element of a clause and can't stand as the first element.
This is an interesting suggestion, which hadn't occurred to me before. On reflection, it seems quite plausible that the odd grammatical myth forbidding initial however was originally a transfer from the treatment of autem and some other words in Latin, beaten into Will Strunk in his youth.
But to start with, the grammar of Latin autem "on the other hand" seems to be a bit of a puzzle in itself.
The most familiar second-position elements in Latin are monosyllabic enclitics like -que "and", -ne "question", -ve "or", which are written as part of the word they attach to; and they can't be clause-initial precisely because they must be attached to the end of some full word. But autem does not seem to be one of these enclitics — it's not a monosyllable, for one thing; it's not written solid with the preceding word; and it's apparently derived from the combination of a connective (which can occur in initial position) with a demonstrative. The Oxford Latin Dictionary analyzes autem as "aut + -em (cf. item and quidem)"; and the Latin connectives et "and", aut "or", vel "or", sed "but", etc., are happy to be clause-initial.
Lewis & Short note about autem that "it is never found at the beginning of a clause, but after one or more words". The OLD says that it is "regularly placed in the second position in its clause, but occasionally postponed, when the first two or three words form a closely-knit (e.g. prepositional) phrase". And indeed, this appears to uniformly true, from early Latin to late, in all the 11,193 instances where this word occurs in extant Latin prose (not that I've checked them all).
Thus in Plautus' Casina we get
Convenit. vin tuis Chalinum huc evocem verbis foras?
tu eum orato, ego autem orabo vilicum.
That's agreed upon. Should you like that, in your name, I should call Chalinus hither out of doors? Do you beg of him, and I'll beg of the bailiff. [Translation by Henry Thomas Riley]
And in Plautus' Menaechmi we get:
Lumbi sedendo, oculi spectando dolent,
manendo medicum, dum se ex opere recipiat.
odiosus tandem vix ab aegrotis venit.
ait se obligasse crus fractum Aesculapio,
Apollini autem brachium. nunc cogito,
utrum me dicam ducere medicum an fabrum.
atque eccum incedit.
My bones ache with sitting, my eyes with watching, while waiting for the Doctor, till he returned from his business. At last the troublesome fellow has with difficulty got away from his patients. He says that he has set a broken leg for Aesculapius and an arm for Apollo. I'm now thinking whether I'm to say that I'm bringing a doctor or a carpenter. But, see, here he comes. [Translation by Henry Thomas Riley]
Or in one of Cicero's orations:
hunc praesentem eis adfecit honoribus quos habuit amplissimos; vos autem absens orat atque obsecrat ut sua religio, laudatio, auctoritas aliquid apud vestros animos momenti habuisse videatur.
They distinguished him while he was among them by the greatest honours which they had to bestow; and now, though absent from this place, they pray and entreat you that their blameless character, their panegyric, and their authority may appear to have some weight with you in forming your opinions. [Translation by C.D. Yonge]
As you can see from these examples, autem is not very much like however in its meaning. Lewis & Short gloss it "on the other hand, but, yet, however, nevertheless; sometimes an emphasized and", and note that "it joins to a preceding thought a new one, either entirely antithetical or simply different".
Because the classical syntax of autem was apparently so uniform, Plautus and Cicero presumably placed it in second position without any explicit instruction — that's just how their language worked. But in the first century A.D., Quintilian's textbook on rhetoric, the Institutio Oratoria, offers explicit advice on this point (I.5.39):
There is, however, some controversy as to the number and nature of the different kinds of solecism. Those who have dealt with the subject most fully make a fourfold division, identical with that which is made in the case of barbarisms: solecisms are brought about by addition, for instance in phrases such as nam enim, de susum, in Alexandriam; by omission, in phrases such as ambulo viam, Aegypto venio, or ne hoc fecit: and by transposition as in quoque ego, enim hoc voluit, autem non habuit.42 Under this last head comes the question whether igitur can be placed first in a sentence: for I note that authors of the first rank disagree on this point, some of them frequently placing it in that position, others never. [Translation by H.E. Butler]
42i.e. nam cannot be coupled with enim; de being a preposition cannot govern an adverb ("from above"); in is not required with Alexandriam, which is the name of a town. Quoque, enim and autem cannot come first in a sentence. Ambulo per viam, ab Aegypto venio, ne hoc quidem fecit would be the correct Latin. [Editor's note from Butler's Loeb edition]
[Yes, it really is footnote 42...]
But enim "indeed" is a bit different from autem, in that Plautus and Terence sometimes use it clause-initially. And quoque "also, even" seems to be be even more different — it's "subjoined to the emphatic word in a clause", as Lewis & Short put it, or "usu. placed directly after the word to which it belongs", as the OLD explains, not placed in second position. Thus in Cicero's Pro C. Rabirio Postvmo:
… me scilicet maxime, sed proxime illum quoque fefellissem.
… I should have behaved dishonestly, principally, indeed, to myself, but in the next degree to him also [Translation C.D. Yonge]
[Though in the particular examples cited above, and a few others that I've now looked at, it could be argued that the word or phrase preceding autem is emphasized as the constrasting element, in a way exactly analogous to the function of quoque. There would still be a difference in the preference for association with emphatic words at the beginning of a clause. ]
Anyhow, the grammar of Latin autem, and the authority of Quintilian in treating the initial placement of autem, enim, quoque as a solecism, may very well be the what put into Will Strunk's mind the germ of the idea that English however ought to be similarly restricted.
Bill Walderman's comment continues:
I don't have access to the OED–-perhaps someone could trace the use of "however" through the history of English to see whether it was used as the first element of a clause early in its history and only later began to be displaced from its position of priority.
The OED's list of citations doesn't really support this idea:
however 3. "Qualifying a sentence or clause as a whole: For all that, nevertheless, notwithstanding; yet; = but at the beginning of the sentence"
1613 SHAKES. Hen. VIII, IV. i. 106 All the Land knowes that: How euer, yet there is no great breach.
1671 MILTON Samson 601, I, however, Must not omit a father's timely care.
1766 GOLDSM. Vic. W. x, This curiosity of theirs, however, was attended with very serious effects.
1790 BURKE Fr. Rev. 27 However, they did not think such bold changes within their commission.
1861 M. PATTISON Ess. (1889) I. 47 It has been even said that this church was built by the Germans, which however was not the case.
1865 LUBBOCK Preh. Times 19 Bronze arrows, however, are not very common in Northern Europe.
Geoff Pullum has speculated that there might have been a late-19th-century fashion for postponing however, but his evidence turned out to be flawed. This idea might still turn out to be statistically valid — and perhaps Latin composition exercises influenced this trend, if this trend existed.
Pending a larger-scale investigation, this whole line of reasoning is speculative at best. But it might help explain where Will Strunk got the strange impulse to declare that Mark Twain used however incorrectly two-thirds of the time.