Corpus lexicographers are used to basing their linguistic judgements on authentic data, on what people have actually said or written. This approach has led to major advances in the study of language, but what it tends to underplay is what people think of their own (or other people’s) language habits, in terms of correctness or acceptability. With this in mind, I recently picked up a short book published in 1970 by Oxford University Press, entitled Attitudes to English Usage. The authors, a team of researchers from the University of Newcastle led by W. H. Mittins, surveyed a total of 457 native English speakers, mainly from the world of education or the liberal professions, on the acceptability – in both spoken and written registers – of an initial set of 50 sentences, each intended to encapsulate a contested point of grammar or usage. Many of the sentences may sound to us today rather quaint (They bought some tomatoes off a barrow-boy.) or entirely unobjectionable (I will be twenty-one tomorrow.), but ‘twas not ever thus.
There seems to have been some recent research interest in Mittins et al’s survey, so I thought it might be useful (in an occasional series) to re-examine some of the sentences in the light of up-to-date corpus data. Let’s start with the most uncontroversial one, achieving the highest percentage of acceptability (86%) from the original respondents:
He did not do as well as the experts had expected.
I imagine it will be quite mystifying to most readers under 40 why anyone might deem this sentence to be incorrect. However, Mittins et al quote a number of authoritative sources, one as late as 1960, to the effect that so should be used instead of the first as in negative comparisons of this kind:
He did not do so well as the experts had expected.
My treasured 1915 copy of Carrad’s English and Commercial Correspondence is certainly pretty categorical on this point, but I’m guessing that most people reading this post would be more likely to use as. All three examples of the as … as construction in the Macmillan English Dictionary (2002) are negative sentences.
Turning now to the corpus data, the massive Google Books corpus shows that not so well as has in fact occurred more frequently than not as well as for most of the past 200 years, finally being overtaken in 1952 (and as late as 1971 for British English considered separately). The crossover seems to occur somewhat earlier in the case of not as / so good as and not as / so bad as. Readers might like to try replacing good and bad with other common adjectives or adverbs, to see if there are any significant differences in timing.
A comparison between the British National Corpus (early 1990s) and ukWaC (2007) shows that as is still gaining ground over so in the pattern not as / so * as, from being 4.17 times more frequent in the early 1990s to 6.21 times more frequent in 2007. In the Corpus of Contemporary American English (1990-2012) the gap is even more pronounced (7.57 times more frequent).
Having talked to many English teachers from many different countries during my career, I am often struck by the persistence of ‘old-fashioned’ rules of grammar and usage in educational syllabuses and textbooks. I would certainly be interested in hearing from anyone, native or non-native speaker, who was taught this rule or attempts to observe it. It may not be so / as old-fashioned as we might think.

I am 27 years old, from California, working at a publishing company in Taiwan. I am currently writing English textbooks for ESL learners and just came across this problem of choosing to between teaching “not so … as” or “not as … as” in our books. I personally would choose to go with “not as … as” but my manager seems to favor “not so … as” because of her experience with the TOEIC testing here.
What are your thoughts? What do you think I should do?
I used to be involved in TOEIC testing and never came across any particular strictures regarding the use of “not as … as”. Can you quote me chapter and verse?
As a corpus linguist of philosophical inclinations, I would propose the following formula: “If Idea X can be expressed by both Usage Y and Usage Z, then, if Y can be shown to be more frequent than Z in expressing X, we have no grounds for deeming Y to be unacceptable and Z acceptable (though both could be acceptable).”
Or, as Joseph Priestley put it rather more elegantly in 1762: “Those who wrote in the language while it was a living one will be accounted the standards of it; and even their imperfections must be adopted by all who use it after them.”
Loving the Priestley quote, John. (I wanted to claim him as a Brummie, but I see that he just lived here, from 1780-1791). I have to say I had never encountered these strictures against ‘not as … as’ and I’m well over 40. The ‘so’ version just strikes me as a little more formal. And in the corpus I consulted (enTenTen) the disparity in terms of numbers was even greater than your numbers from 2007. So it seems like ‘so’ is on the way out.
I had “as well, but not so well” drummed into me in school in the ’60s, and have been using it ever since. I’m willing to change. I just need someone to blame. I do proofreading for friends.
One of my earliest usage memories–so with the negative comparison. Using it tends to emphasize the negative. Learned grammar and usage in the US in the late 50’s and early 60’s. Private Catholic school.
My other recollection: use of the gerund where it is modified by an adjective especially a personal adjective as in:
I appreciate his coming to visit (not him coming).
I am a free-lance writer 85 years of age. I was emphatically taught that the construction should be “not so well as.” This is so ingrained that if I should slip and write “not as well as” alarms go off in my head and immediate correction occurs.
Thank you for your comment, David. I think many of us stick to what we were taught at school, because as you say these early lessons are deeply ingrained.
In the example “He did not do as well as the experts had expected”
His result and the experts’ expectation should be weighed.
If the experts predicted 95% and he scored 90% then:
“He did not do as well as the experts had expected”
If the experts predicted a bad score then
“He did not do so well as the experts had expected”
There is very different meaning attached to “as well” vs “so well”
Hi Julius,
Thanks for your comment. I’ve not heard that distinction before but it would be interesting to test it against corpus data. I could set that as a task for my lexicology undergraduates.
John