I put the spelling 'filosophy' in a Google Ngram and got this very strange chart.

It seems to me that misspellings and grammatical mistakes will always be present in texts, if only as printing errors, so what level of usage should we consider to below the 'noise floor' (that is, just random) in an ngram chart? The question is particularly significant when usage changes over time. At what point can we consider a usage to have died out, even if it turns up faintly on an ngram?

Oh, and why doesn't the word 'jabberwocky' show up on an ngram till 1890, when "Through The Looking Glass" was published in 1871?

  • 3
    I am not quite sure how to interpret your chart, and your question. So you are asking about how serious an issue is, if there is a typo in, say, one-in-a-million books?
    – Matsmath
    Aug 21, 2016 at 11:04
  • @Matsmath, I'm asking what level of incidence in an ngram we can assume to be purely about typos and other errors. Do you understand the term 'noise floor'? 'Noise' in this sense basically means randomness. Is one-in-a-million indicative of randomness? A lot of literary works coin original phrases that have never been used before.
    – Dunsanist
    Aug 21, 2016 at 11:11
  • Looking at the books cited, it seems that the term File Filosophy was intentional.
    – Lawrence
    Aug 21, 2016 at 11:14
  • 'The Love Song Of J. Alfred Prufrock', one of the most famous poems in English, was first published in 1915. However, 'Prufrock' doesn't show up in the ngram till 1931. I'm willing to believe the big gap in the ngram is due to WW2, and shortage of paper, but why does 'Prufrock' only kick in in 1931? books.google.com/ngrams/… I'm quoting this instance because 'Prufrock' is definitely a word T.S.Eliot invented. So there is no noise for it before that.
    – Dunsanist
    Aug 21, 2016 at 11:14
  • @Lawrence aha, so the ngram seems not to be random! So does this mean the 'noise floor' would have to be set lower than that?
    – Dunsanist
    Aug 21, 2016 at 11:16
  • Filosofy is a plausible misspelling of philosophy, whereas Prufrock is not a plausible misspelling of anything. The noise floor is highly word-dependent. Aug 21, 2016 at 11:16
  • From Ngram:.books.google.it/…
    – user66974
    Aug 21, 2016 at 11:17
  • 1
    Ngram already has a noise floor. Which is probably why "Jabberwocky" doesn't show up until 1890. Aug 21, 2016 at 11:20
  • 1
    @Dunsanist: Go look at the online FAQ. Why are you showing a 0% flatline when I know the phrase in my query occurred in at least one book? Under heavy load, the Ngram Viewer will sometimes return a flatline; reload to confirm that there are actually no hits for the phrase. Also, we only consider ngrams that occur in at least 40 books. Otherwise the dataset would balloon in size and we wouldn't be able to offer them all. So that noise floor was set for technical, and not statistical, reasons. Aug 21, 2016 at 11:24
  • 1
    books.google.com/ngrams/… which spelling would you prefer, and why? 1914 Prufrock appears to be also a surname. When using Ngram you should look at the results and also check there are no misspellings in the OCR reader.
    – Mari-Lou A
    Aug 21, 2016 at 11:47
  • 1
    But you argued that Prufrock only appears in 1931, (but why does 'Prufrock' only kick in in 1931?) link It's simply not true. Here's an Ngram showing that Jabberwocky appears around 1872
    – Mari-Lou A
    Aug 21, 2016 at 12:03
  • 1
    From wordoriginsorg.yuku.com/topic/8931/… A letter of enquiry in regard to J. Alfred Prufrock's origin was sent to Mr. Eliot and elicited the following reply: 'Several correspondents have recently called my attention to the Prufrock-Littau company, furniture dealers of St. Louis. I did not have, at the time of writing the poem, and have not yet recovered, any recollection of having acquired the name in this way, but I think that it must be assumed that I did, and that the memory has been obliterated.'
    – Dunsanist
    Aug 21, 2016 at 12:10
  • 1
    OK, good to know. So what is your question? Ngram records variants of spelling? If a user wants to search for a proper noun they should use a capital letter?
    – Mari-Lou A
    Aug 21, 2016 at 12:13
  • 2
    This is actually a statistics question and not an English language question. And it's not going to have a simple answer, like you seem to expect it to. Aug 21, 2016 at 13:25
  • 1
    This is at least a matter for moving to meta, because it's about a tool EL&U posters frequently use. Methodology papers appear in scientific journals.
    – Spencer
    Aug 21, 2016 at 21:05

1 Answer 1


By way of confirming Peter Shor's observation about Ngram's default case-sensitivity, here is the Ngram chart for "Jabberwocky" (blue line) and "jabberwocky" (red line) for the period 1800–2000 (with smoothing reduced to 0 from 3):

The earliest match that the Google Books search results listed in the links beneath this graph find is from the June 15, 1872, in Once a Week; and the next-earliest match is from the December 1882 issue of St. Nicholas. In both instances, Jabberwocky is initial-capped.

If you reduce the time period from 1800–2000 to 1870–1900, you get a magnified image of the relevant segment from the first chart:

But in the links beneath this graph, you'll find nine unique, previewable matches from the period 1870–1879 alone, including an 1875 edition of Through the Looking-glass: And What Alice Found There, which very conveniently details in its front matter the moves in a chess game corresponding to Alice's adventures, in which "White Pawn (Alice) to play, and win in eleven moves."

The main point here is that the degree of detail provided in the Ngram-associated Google Books search results varies tremendously depending on what time interval you use, but the Ngram line graph seems quite stable (aside from predictable changes in magnification).

As for a baseline noise floor, much depends on the years over which you are running your search. Searches going back to the 1600s produce many OCR errors based on blurred, faded, or broken page images; old-fashioned font sets (especially in connection with italics); foreign words misread as English ones (especially if Latin texts are proportionately much more common in the results); books and journals with multiple columns per page and narrow spaces between them, which the OCR sometimes jumps incorrectly; and (in particular) words that use the old-fashioned lowercase "long s" (ſ), which Google's OCR system often misreads as a lowercase F (f).

As you get closer to the present, image quality generally improves and the incidence of OCR errors drops considerably. By the middle-1800s, Ngram/Google Books OCR accuracy is much better than for the early 1800s and before. Nevertheless, searches that bring up disproportionately many newsprint articles have a higher noise problem, due to small type, smudging, and fading of the type.

So my conclusion is that there is no consistent baseline or floor to Ngram noise. Every search must be appraised and fine-tuned on its own unique terms. But the only way to get a sense of what results to trust is to go into the search result links and check each match.

The earliest match for filosophy, for example (and the one responsible for the skyscraper centered on the year 1820 in your link), is to a page in an 1820 Parliamentary history that got creased, causing successive lines to lose the equivalent of 1½ visible characters; hence the OCR misreading. But matches for filosophy from the late 1800s are not errors—they are matches to publications that evidently took a half-baked approach to spelling simplification and balked at filosofy.

  • That all makes sense. However: this shows the incidence of the phrase 'I was rejoiced to' which I found in an Edgar Allan Poe text. Plainly the usage has declined--when should we declare it dead or obsolete? There's still plainly a noise floor there.
    – Dunsanist
    Aug 25, 2016 at 11:28
  • Oh, and there's the problem of reprints. Does Google ngram categorise reprints under the original publication date or the reprint date? Does the 'I was rejoiced to' expression still come up because of E.A.Poe reprints?
    – Dunsanist
    Aug 25, 2016 at 11:32
  • @Dunsanist: The publication date in every instance is the year of publication listed at the front of the book—which means that anything that appears in The Pilgrim's Progress or Gulliver's Travels, say, has produced matches in countless reprints with different publishing dates across the years and continues to do so today. Another problem is that certain publishers (most vividly, to my recollection, Harper & Brothers) prominently display on their title page the year when they began publishing; Google Books' OCR sometimes picks that date up as the date of publication of the particular book.
    – Sven Yargs
    Aug 25, 2016 at 15:28
  • Ha! So Google Ngram is pretty useless then--it can only give the most general trends...
    – Dunsanist
    Aug 27, 2016 at 11:37
  • Ngram does some things very well: it reflects broad changes in frequency of published phrases; it shows roughly when a term came into use over the past 300 years; and it gives a pretty clear indication of which of two alternative wordings has been more common over time. The more common and yet narrowly understood a word or phrase is, the more useful its representation is. ...
    – Sven Yargs
    Aug 27, 2016 at 17:36
  • For example, the Ngram for "bell the cat" for the period 1650–1900 suggests that term was not widely used in English writing before 1700—but search results below the graph include links identifying it as a Scottish proverb as early as 1678. ...
    – Sven Yargs
    Aug 27, 2016 at 17:57
  • ... So working from Ngram, you can identify a very early published occurrence of the expression and get a strong indication of where it as first popular in speech and get a sense of when it caught on in English writing as an idiomatic phrase (sometime after 1700). All of that seems useful to me. Ngram doesn't faithfully represent every published occurrence of a word or phrase—but it doesn't claim to. There is a vast territory between "faithfully represents every published occurrence" and "only gives the most general trends," and Ngram inhabits that territory.
    – Sven Yargs
    Aug 27, 2016 at 18:08
  • Okay, so it seems to me that what ngram is charting is a bit like a musical sound. There's the origin of the sound, there's the question of sustain…and then there's echoes and reverberation. The echoes and reverberation will blur the tail of the sound. If the sound (or usage) stays at the same level, that blurring is irrelevant. Ditto if usage increases. But if usage dies away, 'echoes' (reprints) become more significant. Which leads me to suspect...
    – Dunsanist
    Aug 29, 2016 at 10:26
  • …that literary usages will show up disproportionately in this way. After all, 'Jabberwocky' is a purely literary word, and wouldn't show up at all except for literary use. So I would say that Google ngrams will tend to overemphasise literary usages: that is, turns of phrase, stylings, and even words, that come purely from creative writing. Contrast that with this for 'phlogiston', a concept that...
    – Dunsanist
    Aug 29, 2016 at 10:35
  • …is now completely outmoded in the physical sciences--any modern occurrences of it are in the context of historical discussion.
    – Dunsanist
    Aug 29, 2016 at 10:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .