Birth of the Cool (not about Miles)

I recently published, together with Olivier Morin, a paper in Cognition and Emotion: Birth of the cool: a two-centuries decline in emotional expression in Anglophone fiction. The main result is about a clear decrease in the emotional tone in English-language literature, starting plausibly from the XIX century, a  decrease driven almost entirely by a decline in the proportion of positive emotion-related words, while the frequency of negative emotion-related words shows little if any decline. In other words, English literature became in the last centuries less “emotional” and, in particular, less “positive”.



There are a few reasons why I am quite proud of this paper. First, it builds directly on my previous works, in particular The Expression of Emotions in 20th Century Books. Second, it directly addresses a (fair) criticism that was moved to the above, and in general to all analyses that are exclusively based on the Google Books Ngram Corpus. To put it briefly, we do not know what is inside the Google corpus, and results that are only based on it could be due to unknown changes in the composition of the sample through years, for example. To deal with the problem we compared the Google Books results with two “small data” corpora that we built by ourselves, using a few hundreds of texts freely available in the Project Gutenberg, for which we have all the metadata needed (all information on these corpora are at The “big” and “small” data results were equivalent.

Third, and probably most importantly, our previous works on the topic were only descriptive, i.e. we were just showing the existence of this “emotional decline”, while in this paper we do an explicit effort (especially thanks to Olivier) to provide some theoretical background to the finding, and ultimately to understand what can explain this decline. We examine several hypothesis. Some, such as demographic dynamics concerning age or gender balance, changes in vocabulary richness, or changes in the prevalence of literary genres, we can exclude by analysing the data we have (in particular our “small data” corpora). We discuss other options, including an actual decline in happiness (implausible), a bigger commercial success of low-toned, pessimistic, works (not really compatible with our data), urbanisation (possible, but some questions remain open), and a simple “regression to the mean” with respect to emotionally exuberant Romantic writers. Much more on all this is in the paper.

Almost at the same time, other papers that confirm or expand our main results have been published. This, a contributed PNAS paper, mostly replicates some of our previous analyses, while this study finds the same emotional decline in a sample of German texts of 18th and 19th century, using a different metric for quantifying emotions (the Valence-Arousal-Dominance dimensions), and adapting historically the affective lexicon to “prior language stages”. In sum, this convergence (including other sources we mention in the paper) suggests that this may be a real phenomenon.

At this point I have plenty of questions and future developments in mind (which probably I will mostly not be able to realise), such as, in no particular order:

  • are there robust ways to test possible explanation for the emotional decline (positive but not negative), including urbanisation, regression to the mean, etc.?
  • is possible to apply more sophisticated, not bag-of-words, models to the same problem (as suggested by Bill)?
  • how does this emotional decline relate with the apparently opposite cultural evolution findings that emotional content enhances memorability and transmissibility of stories?
  • what are the specific words that drive the dynamic, and can understanding this be helpful to understand the reason behind it (just had a quick twitter exchange about this with a student in the  Computational story Lab)?
  • what is behind the asymmetry positive/negative emotion?
  • what happen in other languages/literary traditions? It looks like the same decline is present in the non-English samples of the Google corpus (see this figure in the paper), but this remains an open question.


Acerbi A., Lampos V., Garnett P., Bentley R. A. (2013), The expression of emotions in 20th century booksPLOS ONE, 8(3), e59030

Buechel S., Hellrich J., Hahn U. (2016), Feelings from the Past – Adapting Affective Lexicons for Historical Emotion Analysis, in: Proceedings of the Workshop on Language Technology Resources and Tools for Digital Humanities.

Iliev R., Hoover J., Dehghani M., Axelrod R. (2016), Linguistic positivity in historical texts reflects dynamic environmental and psychological factors, PNAS, 113(49), E7871-E7879

Morin O., Acerbi A. (2016), Birth of the cool: a two-centuries decline in emotional expression in Anglophone fictionCognition and Emotion1-13


4 thoughts on “Birth of the Cool (not about Miles)

  1. A few comments:

    I think the phenomenon is real. A number of digital humanists have been finding directionality of various kinds in historical data, including 19th century Anglophone literature. I note, however, that most of them seem to have little or no interest in cultural evolution. I have a number of blog posts about CE directionality.
    In particular, Matthew Jockers has analyzed a corpus of 3300 19C Anglophone novels in Macroanalysis: Digital Methods & Literary History (2013). His last study looks for influence between authors. He operationalizes influenze by calculating Euclidean distance between the texts (in 600-dimensional space) and then visualizing a graph that retains only links between highly similar texts. While the database doesn’t contain any dates, the visualization clearly shows temporal direction. Jockers explicitly rejects cultural evolution as a mode of thought. I of course do not, and I have argued (at some length) that his result indicates directionality for his corpus.
    As to why…we’ve got lots to think about, don’t we? Note however that you are looking at emotional valence in works of fiction, and that is not the only thing in those works. The emotional valence of a work must surely be consistent with whatever else is going on. And surely the memorability and transmissibility of these stories is a function of the entire story, not just some aspect of it taken in isolation. Does this mean we have to look at the aesthetics of these texts? If so, we’re not going to find it by examining them feature by feature. We’re going to have to think about the coherence of the features among themselves.

    1. Hi Bill,

      Thank you for your comments. I agree, emotional valence is just one of the aspects, and to try and understand the “big” trend one needs to consider how various features interact with each other. One has to start somewhere though…Given that we are using new methodologies in a relatively new field, it is probably good having some robust result, even if can not explain the entire picture. Of course I look forward to other contributions!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s