Why Are (Some) Tweets Getting Shorter?

By Neuroskeptic | November 11, 2013 2:34 pm

I like Twitter and I like scientific papers.

So I like this new paper by University of the Philippines researchers Christian M. Alis and May T. Lim: Spatiotemporal variation of conversational utterances on Twitter

Using Twitter’s API, they downloaded 229 million ‘conversational’ tweets from 2009-2012. They defined as ‘conversational’ any tweet starting with the character @. These are messages directed at one or more specific users, although anyone can read them.

The headline finding was that @ tweets are steadily getting shorter, as can be seen in the graph on the right showing linear downward trends over the three quartiles of the distribution:


Shorter how? The difference is mainly due to people using fewer words. The length of the most-used words didn’t change very much, but the number of words per tweet fell:


So tweeters are becoming less verbose (within any given tweet), which the authors suggest might represent the development of more economical linguistic conventions adapted to Twitter. But is this true of everyone?

Broadly speaking, yes – at least in terms of English-language tweets. The slope of the decline was similar in the USA and in tweets originating from the rest of the world.

However within the US, Alis and Lim found a remarkable state-by-state variability (Bear in mind however that few tweets have geolocatable info, so the sample sizes,and representativeness, of these data here are lower)

state_tweetsThe average @ utterence from Louisiana is just 27 characters, compared to 43 in Montana.

Why? State average income and educational attainment were weak predictors of length, but Alis and Lim say that the biggest factor they found was… race. States with more African-Americans produced shorter tweets.

The authors say that

A possible explanation is that Blacks converse more distinctly and more characteristically than other racial groups.

Since utterances were only weakly correlated with income and education then perhaps the shorter utterance lengths is a characteristic of their race – perhaps pointing towards the controversial language of Ebonics [26].

The strong correlation does not imply causality however…


ResearchBlogging.orgAlis CM, & Lim MT (2013). Spatio-temporal variation of conversational utterances on twitter. PloS one, 8 (10) PMID: 24204968

  • http://againstjebelallawz.wordpress.com/ Enopoletus Harding

    The average


    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      The median is, indeed, one kind of average, and the one I was referring to (hence why every graph I showed had “median” on it.)

      • http://againstjebelallawz.wordpress.com/ Enopoletus Harding

        It is, but, since most of the time “the average” is used to mean “the mean” I find use of “the average” to refer to the median somewhat confusing.

  • http://www.facebook.com/people/Joey-Boots/690621350 Joey Boots

    Whoa this is racist – i dunno why but it is!

  • donniagw595

    My Uncle Declan just
    got Fiat 500 C Abarth by working part time off of a macbook air… Look At This

  • Murilo Giacometti

    I thought this was pretty interesting… until I read the conclusion. They should check better their hypotheses before spitting them out when it comes to race; otherwise, it just sounds like prejudice.

  • Mike

    let the leftwing whining commence…

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Whine? Whites should – brevity’s good.

  • citizeny

    I don’t know what’s so controversial about the fact that many African-American people talk differently. It’s a well-documented fact (https://en.wikipedia.org/wiki/African_American_Vernacular_English ) but one need not read Wiki to see it.

    Actually, in my opinion it’s the best dialect/variant of English and now we got proof that it appears to be the most efficient too.

    Btw, how did the authors get data on race, income or education?

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      The data were US state level. % African-American in each state predicted state median tweet length. They didn’t have info (and it would be hard to find it afaik) on individuals.

  • Final_Word

    Ebonic tweets are shorter.
    Fo sho.

  • Ramond Gonzalaz

    This study was done poorly but at the very least they did say :
    Alis CM, & Lim MT (2013)”correlation does not imply causality however”….

  • Tracy L. K.

    And….Why exactly is this important information & why did someone waste their time doing it? Why did we need to know this feeble nonsense? I knew exactly what race it was going to be. I won’t go into how the english language has been butchered by said race either. When you can’t even pronounce simple words such as “library – libary” & “ask – axe”, of course your going to have shorter tweets! Give me a break! It doesn’t take a rocket science to figure that out! Fo Sho!?? Pardon me now while I go throw up!

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Ahem. I see you’re from Texas, home of “fixin to” and “y’all”.

      • Tracy L. K.

        No, I am not from Texas! I Don’t talk like that & neither do the people I know who actually finished high school, went on to prominent colleges & learned how to read & write proper english! When you have illegals coming over here that cannot even speak the language & a large % of blacks that cannot even pronounce the majority of words in the English language let alone understand the definitions, it won’t be long before our language is unrecognizable all together!

  • Kaplan

    Did this study take into account the uneven growth of mobile tweeting? At minimum, which race (ugh)/social/economic subgroups were tweeting by which platforms, where and when? If not, the resulting “conversation” data is meaningless. Fon twts n txts r shrt, as everyone knows, for reasons that have nothing to do with language or race. Obviously, the “platform factor” has to be controlled for in the data before trying to draw any wider conclusions.

    Txtspk’s influence on language and communication outside of the texting environ is interesting, I guess, but isn’t what this study looks at and has been talked to death anyway.

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      Of course. Why didn’t I think of that?

      That’s a most plausible explanation, and the authors didn’t address it (perhaps their data didn’t allow them to.)

  • Pingback: Morsels For The Mind – 15/11/2013 › Six Incredible Things Before Breakfast()

  • http://petrossa.me/ petrossa

    Tweets contain mostly babbling teenagers and housewives, and they use shorthand to be faster on their mobile. Pointless study in kicking open doors in.

  • Pingback: Brevity Is The Soul Of Twitter | The Penn Ave Post()

  • Pingback: day in city | Tweets getting ever shorter()



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar