Back in 2005, Mark Liberman and I (here and here and here) both took a look at certain issues relating to placement of clause adjuncts, and we touched on William Strunk's prejudice against sentence-initial however as an adjunct, as set forth in The Elements of Style. I suggested in "Fossilized prejudices about however" that Strunk had some basis for his prejudices, since novels of the time really did seem to prefer however in second position. This was a modest defense of Strunk, whose horrid little book I regard as almost entirely mistaken in the grammatical advice it purveys. Michael Stillwell has now discovered that my defense evidence was flagrantly mistaken.
I asserted that in Bram Stoker's Dracula (1897) there were 79 cases of ", however," — indicating a second-position use — but no occurrences at all of the string "However," — which is what you would find in a case of the adverb in the sentence-initial position that Strunk advises against. But it has been pointed out to me by Michael Stillwell that (there is no other way to put it) mistakes were made. He gets 18 occurrences of "However," in Dracula; and on re-checking my own file of the text of that novel I get 17 (we must have slightly differing files of the text). Clearly I mistyped something when I did my count in 2005, and looked for a pattern that never occurred. So my erroneous result must be, and hereby is, retracted. Dracula confirms what I would have expected: that the adverb however was perfectly normal in sentence-initial position in novels published when Strunk was a young man. A sixth of Stoker's uses of adverb however are sentence-initial.
Whatever error I made in the search command on Dracula, I then apparently repeated on Joseph Conrad's Heart of Darkness (1902), perhaps by re-using a command line. The figure of 3 cases of second-position however was correct, but although I said there were no cases of initial-position However, that was another mistake: the figure is 8. So Conrad's prose, dating from the time when Strunk was in the early stages of his academic career, clearly disconfirms his style prejudice: 73 percent of Conrad's occurrences however in that novel are sentence-initial.
I also messed up the search on Jack London's The Call of the Wild (1903): I said it was 4 second-position and none sentence-initial; the correct figures are 5 second-position and 1 sentence-initial.
The only novel I looked at in 2005 where my count was right was H. G. Wells's The War of the Worlds (1898). Re-checking the text of Wells's novel confirms that it has 10 second-position occurrences of however but none that are sentence-initial, so it seems to be written in a way that conforms to (or perhaps forms part of the basis for) Strunk's prejudice.
What are the lessons here? I see three:
- Results in linguistics need to be re-checked, and attempts at replication, even of apparently very simple things like counts of word sequences in familiar literary works, will sometimes be very useful. People make mistakes. And I am people too. Many thanks to Michael Stillwell for discovering my error. And apologies to Language Log readers who trusted me to get things right. I goofed.
- We need a larger corpus of results on placement of however in prose of the half century between 1869 (when Strunk was born) and 1919 (when he taught English grammar at Cornell to the man who was later to revise The Elements of Style, E. B. White) if we are to figure out the degree of textual support for Strunk: his edict currently looks highly implausible either as grammar rule or as style recommendation.
- Contrary to what I said in 2005, the present state of play is that so far there is no support for Strunk's rule except that H. G. Wells did avoid any sentence-initial howevers in at least one of his novels. Research continues.