6

I am curious about how often the word gaslighting (in the contemporary sense of "a form of manipulation through persistent denial, misdirection, contradiction, and lying in an attempt to destabilize and delegitimize a target" — Wikipedia as represented in TFD) is being used compared to other years in the preceding century. It seems to me that it might be on the uptick in the last decade, but I have no evidence that it is.

I am particularly interested if the question can be answered using an online resource similar to Google Books Ngram Viewer. I'm hoping for an answer that uses an online resource or resources to show authoritative frequency statistics, year-by-year, for use of gaslighting in journals, popular magazines, newspapers, or online counterparts.

I would especially like to know how much more often gaslighting is used in the last couple years compared to 20 years ago and 50 years ago. Before 1944, I would expect that "gaslighting" would be about lighting a room with liquid propane or natural gas.

migrated from english.stackexchange.com Aug 2 '17 at 7:56

This question came from our site for linguists, etymologists, and serious English language enthusiasts.

  • 1
    No, nothing like that currently exists. COCA and BNC allow filtering by source document (newspaper vs journal vs book vs spoken language) but don't sort or filter by data of occurrence. – Mitch Jul 31 '17 at 14:09
  • I'm voting to migrate this to Meta, since it's requesting resource advice. Nothing against your question Robert, but questions requesting advice on resources tend to be closed as "too broad" on the main site. – RaceYouAnytime Jul 31 '17 at 16:52
  • trends.google.com/trends/… this is interesting, you can trace how often the term gaslighting has been searched in the UK, or anywhere else in the world. And you can search as far back as 2004. This is the result for the UK. – Mari-Lou A Jul 31 '17 at 19:29
  • Between the 22 and 28 January 2017, there was a peak in the number of searches in the US, more or less the same time when DT officially became POTUS. Are the two connected? No idea. – Mari-Lou A Jul 31 '17 at 19:34
  • And Google confirms, gaslighting and Trump were indeed matched. – Mari-Lou A Jul 31 '17 at 19:38
  • @Mari-LouA - this is interesting. You know what the source is? Books or also papers, magazines etc.? – user66974 Aug 1 '17 at 14:22
  • @Josh as I understand it, Google trends is a visual report in the frequency a term or expression is searched via their engines. In other words, Google is watching and recording every single term you type into their search engine. But my understanding might be flawed, I haven't looked into it in any detail. – Mari-Lou A Aug 1 '17 at 18:22
  • 1
    @Tonepoet I voted to migrate based on this discussion: What sort of resource request questions do we accept on meta. Word frequency resources seem like something that would be relevant to users of this site on etymology or phrase-origin questions. Maybe I was wrong and this isn't a proper candidate, but when I encountered it, it already had upvotes, upvoted answers and no close votes. I thought migration was a reasonable vote. – RaceYouAnytime Aug 2 '17 at 13:15
  • @Tonepoet (Also, my vote was before the revisions, it was much more clearly a resource request at the time) – RaceYouAnytime Aug 2 '17 at 13:20
  • 1
    @RaceYouAnytime I understand. We have a long standing history of this. I just pinged you mostly for your future reference, and partially so that you'd know what I was talking about in the other discussion, since I mentioned it there. I don't take it that even a plain resource request would be allowed here with such broad wording though. – Tonepoet Aug 2 '17 at 13:33
  • @Tonepoet that's reasonable, point taken. – RaceYouAnytime Aug 2 '17 at 13:34
5

It is not entirely clear from your question whether your problem with using Google Ngrams is that the data is not recent enough or that it is based on formal writing (books only), or both. But there are a number of other publicly accessible corpora that might suit your needs:

The Corpus of Contemporary American English has data from 1990 to 2015, and includes transcripted speech (from TV and radio), popular magazines, newspapers, and journals as well as fiction.

The Corpus of American Soap Operas is based on a smaller data set but runs from 2001 to 2012 and is good for exploring more colloquial/informal speech, albeit scripted rather than real-world.

The News on the Web (NOW) corpus is a very large corpus of some 5 billion words covering roughly the last decade of online news reports. It can be downloaded in its entirety free of charge.

And there is also a more or less up-to-date Wikipedia Corpus based on around 2 billion words from the 4+ million articles of the famous encyclopaedia website.

  • the question is simply as stated. if such exists, i would like to use something very similar to the Google Books Ngram Viewer to research the relative use of a word in newspapers, magazines, journals, and online publications. i wanna know how much more often is this word referred to in the last couple years compared to 20 years ago. and then 50 years ago. the movie was about 75 years ago. so before 1944, i would expect that "gaslighting" would be about lighting a room with liquid propane or natural gas. now, when the word is used, it is disparagingly used against someone in an argument. – robert bristow-johnson Jul 31 '17 at 5:03
  • 1
    @robertbristow-johnson, you should edit your question to include this information, as users will not necessarily see this comment before attempt to answer. – vpn Jul 31 '17 at 5:06
  • 1
    @robertbristow-johnson Thanks for the clarification. I understood what you were trying to achieve but not why you felt Google Ngrams wasn't an adequate tool for the job. – Daniel Austin Jul 31 '17 at 5:09
  • because it's just monograph books. not magazines nor journals nor newspapers. – robert bristow-johnson Jul 31 '17 at 5:11
  • @vpn , how would i edit the question differently? the comment has nothing more than the original question. BTW, you or anyone else with edit capabilities are welcome to edit the question to make it more clear in your estimation. if i don't think it helps, i might revert or further edit it. – robert bristow-johnson Jul 31 '17 at 5:14
  • Daniel, thank you for pointing me to Wikipedia Corpus. i will make use of it in the future. for the kind of question i want to answer, it does not tell me what i am looking for. but thank you nonetheless. – robert bristow-johnson Jul 31 '17 at 5:23
  • @robertbristow-johnson Yes, those resources are good for investigating recent trends but I appreciate they are limited in how much they could help you to compare the status quo to several decades ago. In any case, I am not aware of any other corpora that are as user-friendly as Google Ngrams in generating charts etc. I think Microsoft tried making a rival corpus ngram at one point but not sure it even exists anymore. – Daniel Austin Jul 31 '17 at 5:28
  • please don't take it personally that no check mark. i did upvote the answer. but so far no one has been able to really answer the question (and there might not yet be an answer, but i sure would have thought there woulda been something like Ngram that included newspapers, magazines, and journals.) – robert bristow-johnson Jul 31 '17 at 6:16
4

Generally speaking, yes, resources do exist for researching historical word frequency (including contemporary frequency). Here, frequency is defined as

The number of times an event or character occurs in a given sample; also (the relative frequency or proportionate frequency), this number expressed as a proportion of the total possible number of occurrences.

OED; plain emphasis mine.

You may not find that the results meet your goals; the precision is rough if it can be called precision at all, and numerous caveats pertain. For the word you've chosen, 'gaslighting', any comparison will be estimates of estimates, and the results far from statistically precise.

For example, 'gaslighting' in the sense of interest to you (the verb with the specific meaning of "a form of manipulation") is first attested with that sense in 1965 (OED). That is barely more than 50 years ago, and it is reasonable to assume that frequency (in the statistical sense given above) at that time would've been quite low.

The tools to use for the comparison, additionally, would need to be used in combination. COCA, COHA and NOW, especially, could be useful to inform an estimate of the relative frequency of use of 'gaslighting' during the time periods you've specified and in the 'genres' of interest ("journals, popular magazines, newspapers, or online counterparts").

Those corpora, and particularly the genre-balanced COCA and COHA, do allow, generally speaking, comparisons of frequency for specific time periods and in particular genres: the genres balanced, however, are only fiction and nonfiction books, magazines and newspapers for COHA, and spoken, fiction, magazine, newspaper and academic publications for COCA. NOW is useful only for online news publications.

Additionally, the time periods covered by COCA (1990-2015), COHA (1810-2009) and NOW (2010-yesterday) would make comparisons a matter of using all three corpora to find an estimated frequency for 50 years ago (COHA), 20 years ago (COHA and COCA), and the "last couple of years" (NOW). Only one genre or, fudging it, possibly two genres, namely news[papers] and magazines, is available in all three corpora.

In the case of 'gaslighting', the general imprecision of any results may be balanced by features of the three corpora that allow examination of uses in context, by period. So, false positives for 'gaslighting', that is uses of 'gaslighting' in other senses than the target sense, can be readily identified and discarded from consideration.

Frequency of 'gaslighting' from 2010-2017 in the NOW corpus:

gaslighting, NOW

There being only 231 occurrences in the NOW corpus works in your favor. Each occurrence can be seen in context by clicking the bar in the chart (or click "SEE ALL TOKENS" to see them all at once). Here are the occurrences from 2017B (the last half of 2017; at present, this is only July):

gaslighting, NOW 2017B

The context reveals all occurrences are used with your target sense; the display also shows exact date and country of use, along with the particular venue.

Similarly, all occurrences from 2017A are in your target sense. Those, 2017A and 2017B were the only sections I examined for the purposes of this answer. Other sections could be examined, however, and false positives could be discounted from the relative frequency as shown on the chart.

Setting aside examination of context to determine sense, and looking back at the frequency chart, it shows that an increase of 0.04 occurrences per million words of sample distinguishes the first and second halves of 2012; an increase of 0.04 is also shown from the last half of 2016 to the first half of 2017. Otherwise, an increase of 0.03 distinguishes the first and second halves of 2016. Altogether, an increase of 0.10 is recorded from the first half of 2015 through the first half of 2017. Such a marked and gathering increase does not occur over any other timespan in the chart.

Comparing the chart from NOW for 2010-2017 with a chart for occurrences from 1990-1997, 20 years ago is quite simple: no occurrences in any genre are recorded. Two occurrences do appear in 1998. Here they are in context:

gaslighting, COCA 1998

One occurrence is in the target sense, and it appeared in a popular magazine (albeit with a pseudo-scientific focus).

As it happens, that one occurrence would be an increase of 0.01 (from 0.0) per million words, because the corpus size was 103.4 million words in 1998:

gaslighting, COCA 1998 chart

The same techniques can be applied to find the relative frequency of use of 'gaslighting' in your target sense 50 years ago (0.0, because it doesn't appear in the corpus), by comparing results from the COHA corpus.

Oveall, what is expressed by comparison of the results from the corpora, which can be cited as the basis of your estimate, is that use of 'gaslighting' in the sense of "a form of manipulation" increased markedly in the popular press, from near zero fifty years ago to 0.01 per million words twenty years ago to 0.1 per million words in the last two years.

You must log in to answer this question.

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