I was using Google Ngrams to investigate whether feminine pronouns have been used more often in English literature over time, relative to masculine pronouns. I was surprised to find that the percentage of all pronouns out of all words being written dropped steadily from 1800 to 1980 and then shot up between 1980 and 2020 in this graph.

Google Ngram

What could explain why the prevalence of all pronouns would change in parallel with each other? Does it reflect real change in usage or is it an artifact of the Ngrams archives?

  • 2
    This might be a question better addressed to Google (only they will know the algorithms that their code uses and the full content of their databases). – KillingTime Sep 24 '20 at 18:46
  • 1
    Suspiciously selective ('He' bucks the trend). It could be that this is an artifact of Google's sampling techniques, though you may have a very valid point. There are perhaps more reliably sampled databases. – Edwin Ashworth Sep 24 '20 at 18:50
  • 1
    What bothers me is that novels and non-fiction are vastly different. – Lambie Sep 24 '20 at 19:19
  • 2
    As corpus sizes grow over time, the relative composition of corpus texts at any given time also changes. Hypothetically, imagine if a lot of publications in 1800 are published newspapers or magazines, but by 1850 or 1900 mass market yellowbacks make up a much larger share of books. Imagine the growth of nonfiction, of children's literature and (later) young adult lit. Imagine new genres - the self-help book. All this is to say, it's impossible to answer your question without knowing how Google's catalogue changes over time. – TaliesinMerlin Sep 24 '20 at 19:20
  • @TaliesinMerlin is Ngrams a 'living' corpus? (Which is an oxymoron I suppose). The question Laurel cites on meta mentions it was one-off in 2008, rather than any updates to Google's catalogue. – marcellothearcane Sep 24 '20 at 21:13
  • 1
    @marcellothearcane Newer corpora are from 2012 and 2019-2020. "Below are descriptions of the corpora that can be searched with the Google Books Ngram Viewer. All corpora were generated in July 2009, July 2012, and February 2020; we will update these corpora as our book scanning continues, and the updated versions will have distinct persistent identifiers." (Google Ngram Info The poster in the other thread is citing a Wikipedia statement that has now been edited, and thus may not be current. – TaliesinMerlin Sep 24 '20 at 21:58
  • 2
    The ratio of pronouns/words goes down when the numerator (pronoun usage) goes down relative to the denominator (word count). This could be totally unrelated to pronoun usage, and better understood as overall wordiness in writing. This topic requires a deep dive into word statistics. – jimm101 Sep 27 '20 at 1:44
  • It's not like this is a 'really interesting phenomenon'. It is really messed up. It really needs to be explained, otherwise all sorts of other inferences taken from Google NGrams is pretty untrustworthy. Roughly a doubling in frequency in 20 years after a very clear decline over 200. (and comparing with other hi frequency words which all stay pretty flat over the same time period). – Mitch Sep 29 '20 at 17:14

This seems to be just weirdness with Google Ngrams, though I’m not sure what specifically is causing it. I don’t trust Google Ngrams: See the discussion of Google Ngrams here for why.

In comparison, looking at the “Chart” in COHA for PRON I get this:

COHA chart

In other words, pronouns are between .65 and .77% of words in the corpus, depending on the year, which isn’t that much variation.

  • BNC chart shows usage by medium, which could be illuminating: i.stack.imgur.com/yLkAq.jpg. What are the chances "spoken" became much more recorded circa 2000 (which would tie in with the spike in the OP's chart)? – marcellothearcane Sep 24 '20 at 21:23
  • 1
    Thanks for the answer! The [discussion of Google Ngrams] you linked referenced an "overabundance of scientific literature." I assume scientific literature is less likely to use pronouns in favor of precise vocabularly, which could explain the steady downhill slope. Then if Ngrams database had an influx of more personal or fiction writing starting in the 90's we have our explanation for the rise at the end. – Paul Martin Sep 24 '20 at 21:58
  • @PaulMartin It strikes me that the sudden increase correlates with the introduction of the internet. – Greybeard Sep 30 '20 at 23:16
  • Just want to flag that while this chart provides evidence against the decline visible in the Google Ngram, its ending in 2000 means it says nothing about the uptick – Unrelated Oct 1 '20 at 15:36
  • @Unrelated You can do the same search in COCA which covers 1994 to last year and you’ll see that there is no uptick in pronoun usage. It’s all ~.65% pronouns much like in the historical corpus. – Laurel Oct 1 '20 at 15:57

You must log in to answer this question.

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