See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344277133

Adoption of Twitter's New Length Limit: Is 280 the New 140?

Preprint · September 2020

CITATIONS READS
0 168

3 authors, including:

Kristina Gligoric
École Polytechnique Fédérale de Lausanne
19 PUBLICATIONS   116 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Kristina Gligoric on 17 September 2020.

The user has requested enhancement of the downloaded file.

 Adoption of Twitter’s New Length Limit: Is 280 the New 140?
Kristina Gligorić Ashton Anderson
EPFL University of Toronto
kristina.gligoric@epfl.ch ashton@cs.toronto.edu
arXiv:2009.07661v1 [cs.SI] 16 Sep 2020
Abstract
In November 2017, Twitter doubled the maximum allowed
tweet length from 140 to 280 characters, a drastic switch on
one of the world’s most influential social media platforms.
In the first long-term study of how the new length limit was
adopted by Twitter users, we ask: Does the effect of the new
length limit resemble that of the old one? Or did the dou-
bling of the limit fundamentally change how Twitter is shaped
by the limited length of posted content? By analyzing Twit-
ter’s publicly available 1% sample over a period of around 3
years, we find that, when the length limit was raised from 140
to 280 characters, the prevalence of tweets around 140 char-
acters dropped immediately, while the prevalence of tweets
around 280 characters rose steadily for about 6 months. De-
spite this rise, tweets approaching the length limit have been
far less frequent after than before the switch. We find widely
different adoption rates across languages and client-device
types. The prevalence of tweets around 140 characters before
the switch in a given language is strongly correlated with the
prevalence of tweets around 280 characters after the switch in
the same language, and very long tweets are vastly more pop-
ular on Web clients than on mobile clients. Moreover, tweets
of around 280 characters after the switch are syntactically and
semantically similar to tweets of around 140 characters be-
fore the switch, manifesting patterns of message squeezing
in both cases. Taken together, these findings suggest that the
new 280-character limit constitutes a new, less intrusive ver-
sion of the old 140-character limit. The length limit remains
an important factor that should be considered in all studies
using Twitter data.
1 Introduction
On 7 November 2017, Twitter suddenly and unexpectedly
increased the maximum allowed tweet length from 140 to
280 characters, thus altering its signature feature. Accord-
ing to Twitter, this change, which we henceforth refer to as
“the switch”, was introduced to give users more space to ex-
press their thoughts, as a disproportionately large fraction of
tweets had been exactly 140 characters long (Rosen 2017a;
Gligorić, Anderson, and West 2018). Understanding the
consequences of the switch is of paramount importance for
social media studies, for two reasons: first, because Twitter
is one of the leading social media platforms (Perrin and An-
derson 2019), with content posted there reaching and affect-
Robert West
EPFL
robert.west@epfl.ch
ing billions of people across the globe; and second, because
the constraints that a medium imposes affect the audience
not only through the content delivered over the medium, but
also through the characteristics of the medium itself, or, in a
mantra coined by Marshall McLuhan (1964), “the medium
is the message”. Thus, anybody in whose research Twitter
plays a role cannot ignore the switch, and must consider how
the new length limit has impacted Twitter as a platform.
Early work that studied Twitter users’ attitudes toward the
new 280-character limit (Rimjhim and Chakraborty 2018)
discovered varying initial reactions ranging from anticipa-
tion, surprise, and joy to anger, disappointment, and sadness.
Early studies also revealed a low initial prevalence of long
tweets immediately following the switch (Perez 2017) and
studied the short-term impact that the length limit had on
linguistic features and engagement (Gligorić, Anderson, and
West 2018; Boot et al. 2019), also in the specific context of
political tweets (Jaidka, Zhou, and Lelkes 2019).
Whereas the above-cited, early studies necessarily had to
consider short-term effects, much less is known about the
long-term effects of the switch. Now, nearly three years af-
ter the switch, the present paper constitutes the first attempt
to bridge this gap with a long-term study spanning several
years. Broadly, our research was guided by the question, Is
280 the new 140? In other words, does the effect of the new
length limit resemble that of the old one, just over a broader
range of potential tweet lengths? Or has the doubling of the
limit fundamentally changed how Twitter is shaped by the
limited length of posted content?
Research questions. Concretely, we address the following
research questions:
RQ1: How did tweet length change in the two years fol-
lowing the switch?
RQ2: How did tweet length change across languages?
RQ3: How did tweet length change across devices (Web
vs. mobile clients vs. automated sources)?
RQ4: What are the syntactic and semantic characteristics
of long vs. short tweets?
Summary of main findings. Using a 1% sample of all
tweets spanning the period from 1 January 2017 to 31 Octo-
ber 2019, we conduct an observational study of tweet length
 Monthly tweet length histogram In a nutshell, although the doubling of the length limit
2017-1 eliminated the drastic disproportion of tweets reaching the
2017-2 maximum length (e.g., 9% of English tweets used to be
2017-3
2017-4 exactly 140 characters long before the switch), our results
2017-5 demonstrate the emergence of a similar, though considerably
2017-6
2017-7 weaker, effect around 280 characters after the switch. We
2017-8 hence answer the guiding question—Is 280 the new 140?—
2017-9 in a nuanced way: 280 is a less intrusive 140. These findings
2017-10
(the switch) 2017-11 have important implications for Twitter-based research, as
2017-12 they show that, although the new limit is “felt less” by users
2018-1
2018-2 than the old limit, 280 characters still constitutes an impact-
2018-3 ful length constraint that shapes the nature of Twitter.
2018-4
2018-5
2018-6 2 Related work
2018-7
2018-8 Twitter communication and supporting features: stud-
2018-9
2018-10 ies of use and emerging conventions. Previous work has
2018-11 extensively studied communication taking place on Twit-
2018-12
2019-1 ter and the specific features that support them, most im-
2019-2 portantly: retweets (Boyd, Golder, and Lotan 2010), hash-
2019-3 tags (Wikström 2014; Page 2012), quotes (Garimella, We-
2019-4
2019-5 ber, and De Choudhury 2016), and emojis (Pavalanathan
2019-6 and Eisenstein 2016). Previous work has also investigated
2019-7
2019-8 linguistic conventions on Twitter, the patterns of their emer-
2019-9 gence (Kooti et al. 2012b; Kooti et al. 2012a), how users
2019-10 align to them in conversations (Doyle, Yurovsky, and Frank
1

70

140

210

280

number of characters
Figure 1: Monthly tweet length histograms, normalized.
over time (RQ1). Once the length limit was doubled and the
old 140-character limit became obsolete, we find a decline
in the prevalence of tweets of exactly or just under 140 char-
acters, and a rise in the prevalence of tweets of exactly or
just under 280 characters, with a smooth adaptation phase
of around 6 months.
Comparing languages (RQ2), we find vastly different lev-
els of adoption patterns. The prevalence of 140 characters
before the switch is strongly correlated with the prevalence
of 280 characters after the switch, indicating that some lan-
guages have an inherent affinity to longer messages. Com-
paring device types (RQ3), tweets above 140 characters are
used more on Web clients, compared to mobile clients. Au-
tomated sources and third-party applications were the slow-
est to adapt. They continue to tweet around 140 characters
disproportionately more often, compared to regular users,
and in general, tend to publish longer tweets.
Finally, we observe that 280-character tweets are syn-
tactically and semantically similar to 140-character tweets
posted before the switch (RQ4). The 280-character tweets
show linguistic fingerprints indicative of “message squeez-
ing”, with inessential parts of speech (e.g., fillers, adverbs,
conjunctions) being relatively less frequent, and essential
parts of speech (e.g., verbs, negations) being relatively more
frequent, compared to shorter tweets. Their usage is ad-
ditionally associated with specific topics, such as Money,
Death, Work, and Religion.
2016), how they diffuse (Centola et al. 2018; Chang 2010),
and how they continuously evolve (Cunha et al. 2011).
Variations in usage and adoption of conventions and lin-
guistic style. Additionally, previous work has studied how
patterns of adoption of these features, as well as the linguis-
tic style used on the platform more broadly, varies across
numerous dimensions (Shapp 2014), including gender (Ciot,
Sonderegger, and Ruths 2013), political leaning (Sylwester
and Purver 2015), age and income (Flekova, Preoţiuc-Pietro,
and Ungar 2016); but also within accounts (Clarke and
Grieve 2019).
Length limit and the impact of message length on success
and linguistic characteristics. Previous work has studied
how the imposed length constraint on Twitter and other mi-
croblogging platforms affects the dialogues and the linguis-
tic style (Zhou and Xu 2019; Jin and Liu 2017), and the suc-
cess of the message measures through the received engage-
ment (Tan, Lee, and Pang 2014; Gligorić, Anderson, and
West 2018; Gligorić, Anderson, and West 2019; Wang and
Greenwood 2020; Wasike 2013). More broadly, philology,
communication, education, and psychology scholars have
investigated conciseness and its benefits in many different
contexts (Laib 1990; Vardi 2000; Sloane 2003).
Message framing on Twitter. Many previous studies have
investigated the question of what wording makes messages
successful in online social media, often formulated as the
task of predicting what makes textual content become popu-
lar (Berger and Milkman 2012; Guerini, Strapparava, and
Özbal 2011; Lamprinidis, Hardt, and Hovy 2018). In the
specific case of Twitter, in addition to characterizing how
language is used on the platform in general (Murthy 2012;
Levinson 2011; Hu, Talamadupula, and Kambhampati 2013;
 Distribution of tweets 10%
Fraction of tweets, n_chars [136,140]
across languages the switch
8%
Japanese
English 6%
Portuguese

Fraction
Spanish 4%
Arabic
Korean 2%
Indonesian
Tagalog
Turkish 0%

2017-1-1
2017-2-1
2017-3-1
2017-4-1
2017-5-1
2017-6-1
2017-7-1
2017-8-1
2017-9-1
2017-10-1
2017-11-1
2017-12-1
2018-1-1
2018-2-1
2018-3-1
2018-4-1
2018-5-1
2018-6-1
2018-7-1
2018-8-1
2018-9-1
2018-10-1
2018-11-1
2018-12-1
2019-1-1
2019-2-1
2019-3-1
2019-4-1
2019-5-1
2019-6-1
2019-7-1
2019-8-1
2019-9-1
2019-10-1
French
Thai
language

Russian
Italian date
German
Polish Figure 3: Daily fraction (indicated with a circle), and 10-day rolling average (solid line) of
Hindi faction of tweets that have between 136 and 140 characters (inclusive).
Persian
Dutch
Haitian Creole Fraction of tweets, n_chars [276,280]
2%
Estonian the switch
Chinese
Urdu 1.5%
Swedish
Fraction

1%
1M
10M
100M

number of tweets 0.5%

Figure 2: Number of origi- 0%

2017-1-1
2017-2-1
2017-3-1
2017-4-1
2017-5-1
2017-6-1
2017-7-1
2017-8-1
2017-9-1
2017-10-1
2017-11-1
2017-12-1
2018-1-1
2018-2-1
2018-3-1
2018-4-1
2018-5-1
2018-6-1
2018-7-1
2018-8-1
2018-9-1
2018-10-1
2018-11-1
2018-12-1
2019-1-1
2019-2-1
2019-3-1
2019-4-1
2019-5-1
2019-6-1
2019-7-1
2019-8-1
2019-9-1
2019-10-1
nal tweets in the 1% sample
posted between January 1st
2019 and October 31st 2019, date
across the 23 studied lan-
guages where the switch hap- Figure 4: Daily fraction (indicated with a circle), and 10-day rolling average (solid line) of
pened (in orange), and did faction of tweets that have between 276 and 280 characters (inclusive). Note the different
not happen (in blue) y-axis scales.

Eisenstein 2013), researchers have investigated the corre- community level change as opposed to user-level behaviors
lation of linguistic signals with the propagation of tweets since user-level information is incomplete (in expectation,
(Artzi, Pantel, and Gamon 2012; Bakshy et al. 2011; Tan, we have 1% of tweets posted by a fixed user).
Lee, and Pang 2014; Pancer and Poole 2016).

3 Data
We use publicly available 1% sample of tweets, spanning the
period between 1 January 2017, and 31 October 2019, avail-
able on the Internet Archive.1 We consider original tweets
(i.e., we discard retweets). There are between 1 and 1.5 mil-
lion daily original tweets. In Figure 2 we show the exact
number in total across languages.
We study 23 biggest languages: three languages where the
switch did not happen: Japanese, Korean, and Chinese, and
twenty where it happened, each language with more than 2M
tweets in total. The switch did not happen in Japanese, Ko-
rean, and Chinese because the 140 characters limit was not
as restrictive as it was in other languages, since more infor-
mation can be conveyed with the same number of characters
(Rosen and Ihara 2017; Rosen 2017b). We note that the data
is sampled at the community level. We stay at describing the
1 https://archive.org/details/twitterstream
Character counting. We carefully count the number of
characters based on the official documentation2 . Tweet
length is counted using the Unicode normalization of the
tweet text. The tweet text is selected from the tweet ob-
ject using displayed text range information, discarding any
retweet tags, and leading @user mentions that are not
counted towards the length limit. The text content of a Tweet
can contain up to 140 characters (or Unicode glyphs) before
the switch, or 280 after the switch. Emoji sequence using
multiple combining glyphs counts as multiple characters.
Glyphs used in Chinese, Japanese or Korean languages are
counted as one character before, and as two characters after
the introduction of the 280 limit. Therefore, a Tweet com-
posed of only CJK text can only have a maximum of 140 of
these types of glyphs after the switch.
 Histogram of tweet lengths
Hindi Urdu Persian Turkish German Swedish Dutch English
6%
4%
2%
0%
Italian Spanish Polish French Arabic Russian Chinese Indonesian
6%
4%
Probability

2%
0%
Estonian Thai Portuguese Korean Tagalog Japanese Haitian Creole
6%
Jan-Oct 2017
4% Jan-Oct 2019

2%
0%
1 70 140 210 280 1 70 140 210 280 1 70 140 210 280 1 70 140 210 280 1 70 140 210 280 1 70 140 210 280 1 70 140 210 280
Number of characters

Figure 5: Tweet length histograms in the 23 studied languages, for the period before the 280 limit was introduced (in red), and
after it was introduced (in blue). The languages are sorted by the prevalence of 140 before the switch.

Hindi (hi) Next, we focus on these two interesting ranges, close to

6% Urdu (ur)
hi Persian (fa) 140 and close to 280, and do a more granular daily analy-
Turkish (tr)
ur
German (de) sis. We monitor the daily fraction of tweets that either hit
Swedish (sv)
Prevalence of 140 before the switch

fa Dutch (nl) the character limit or are within a five-character margin of

4% English (en)
tr de Italian (it) it. The daily fraction of tweets that have between 136 and
Spanish (es)
nl sv
en
es it Polish (pl) 140 characters are shown in Figure 3, and the daily fraction
frpl French (fr) of tweets between 276 and 280 characters in Figure 4. Here
ar Arabic (ar)
2% zh Russian (ru) we consider again original tweets across the 20 studied lan-
ru Chinese (zh)
in Indonesian (in) guages, where the 280 character limit was introduced. Gray
thpt et Estonian (et)
ko
tl Thai (th) stripes indicate omitted days with missing data.
ja ht Portuguese (pt)
0%
Korean (ko) We note that while 140-character tweets became less
Tagalog (tl)
Japanese (ja) prevalent after the switch, dropping from 7.0% to 1.2% over
0% 0.3% 0.6% 0.9% 1.2% Haitian Creole (ht)
Prevalence of 280 after the switch the studied period, the prevalence of 280-character tweets
steadily increased for 6 months after its introduction, reach-
Figure 6: Fraction of tweets that are exactly 280 characters ing 1.5% at the end of the studied period.
long after the switch, on x-axis, and fraction of tweets that
are exactly 140 characters long before the switch, on y-axis 4.2 RQ2: How did tweet length change across
(Spearman rank correlation 0.92, p = 4.87 ∗ 10−10 ). languages?
Next, we seek to characterize the adoption across the studied
languages. We start by examining in Figure 5 histograms of
4 Results tweet lengths in red Jan-Oct 2017 as pre-switch period, and
4.1 RQ1: How did tweet length change in the two in blue the same months 2 years later when the things settled
years following the switch? in Jan-Oct 2019, as post-switch period. Here we restrict our-
selves to tweets posted from regular sources (web and mo-
First, we measure the prevalence of different tweet lengths bile interface, rather than third-party applications and auto-
over time. We start by inspecting monthly histograms of mated sources). Similar to the overall view, across languages
tweet lengths, shown in Figure 1, where across 20 studied the first peak and the mode of the distribution are constant,
languages we visualize the distribution in red for the months and the interesting character length ranges are near 140 and
before the switch, and in blue for the months after. near 280 characters.
We observe a sharp decline of 140-character tweets and In Figure 6 we show the fraction of tweets that are exactly
an increase in 280-character tweets after the switch. Other- 280 characters long after the switch, on the x-axis, and the
wise, the first peak, consistently between 25 and 30 charac- fraction of tweets that are exactly 140 characters long before
ters, remains unchanged (i.e., the mode of the distribution is the switch, on the y-axis.
stable). We observe that the prevalence of 140 before the switch
in a language is correlated with the prevalence of 280 after.
2 https://developer.twitter.com/en/docs/counting-characters The more 140 was used in a language before the switch, the
 Fraction of tweets n_chars [276,280]
Hindi Urdu Persian Turkish German Swedish Dutch English
6%

4%

2%

0%
Italian Spanish Polish French Arabic Russian Chinese Indonesian
6%

4%
Fraction

2%

0%
Estonian Thai Portuguese Korean Tagalog Japanese Haitian Creole
6%
the switch
4%

2%

0%
2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1

2017-1-1
2017-6-1
2017-11-1
2018-4-1
2018-9-1
2019-2-1
2019-7-1
date

Figure 7: Daily evolution of 280 limit adoption across languages. Daily fraction (indicated with a circle), and 10-day rolling
average (solid line) of faction of tweets that have between 276 and 280 characters (inclusive). The languages are sorted by
prevalence of 140 before the switch.

more 280 is used after (Spearman rank correlation 0.92, p =
4.87 ∗ 10−10 ). In Hindi and Urdu, 280 is very prevalent, and
it is the mode of the distribution–the most frequent character
length after the switch.
In Figure 7, we further monitor the evolution of adoption
patterns of the new limit across languages, for tweets posted
from web and mobile sources. In most of the languages,
the prevalence seems to have settled, and is even decreasing
again, i.e., the peak of usage is past. Urdu is a notable excep-
tion, where the prevalence is still growing, and the adoption
rate is still not in a stable state.
Differences in differences estimation of the effect of
switch. Additionally, we take advantage of the fact that the
new limit was not introduced in all languages to perform
a differences in differences estimation of the effect of the
switch on tweet lengths.
To go beyond visual inspection and to account for pos-
sible global platform-wide changes that are not associated
with the switch, we use a differences in differences regres-
sion estimation, where the tweet lengths in Japanese, Korean
and Chinese, languages where the 280 characters were not
introduced are the control timeseries, and the tweet lengths
in the other 20 studied languages (Figure 2) are the treated
timeseries. Both are observed in the pre-switch (Jan-Oct
2017), and post-switch (Jan-Oct 2019) periods, as illustrated
in Figure 8. We fit a model
y ∼ treated ∗ period, (1)
where the dependent variable y is the logarithm of the aver-
age tweet length for each studied calendar day, and as inde-
pendent variables are the following two factors: treated
(indicates whether the switch was introduced or not in
those languages), period (indicates whether a calendar day
is in year pre-switch or post-switch). treated ∗ period
is shorthand notation for α + βtreated + γperiod +
δtreated : period + , where in turn treated : period
stands for the interaction of treated and period.
The interaction term treated : period δ is then the ef-
fect of switch on the logarithm of average tweet length. Each
studied pre- or post-switch period spans 277 days per condi-
tion, amounting to a total of 4 × 277 = 1108 datapoints. The
model is multiplicative due to the log. The relative increase
over the baseline is then calculated by converting back to
the linear scale the fitted coefficient δ. Fitting the model 1,
we measure a eδ − 1 = e0.0598 − 1 = 6.16% (95% CI [5.68%,
6.64%]) increase in tweet lengths in the languages where the
switch happened, over the control baseline.
To conclude, the estimate a significant increase in tweet
lengths in languages where the switch happened, compared
to the control languages.
4.3 RQ3: How did tweet length change across
devices?
To understand the adoption of the new limit across differ-
ent devices, we monitor the evolution of the daily fraction
of tweets in interesting tweet lengths separately across web,
mobile, and automated sources and third-party applications,
in Figure 9. The tweets are tweeted in the 20 languages
where the switch happened.
We observe different adoption patterns between web and
mobile. Longer tweets are used more on the web. In the web
interface, where 140 was most prevalent (around 12%), 140
was quickly surpassed by 280, reaching around 4% at the
end of the studied period. Automated sources and third-party
applications the slowest to adapt.
Tweet length of 280 characters and the near long lengths
surpassed 140 around June 2018 ( 7 months after the switch)
for mobile, but for web immediately after the switch.
 80 Mobile
Fraction of tweets, n_chars [136,140]
12% Fraction of tweets, n_chars [276,280]
70 the switch
average tweet length 8%

Faction
60 280 introduced
(20 languages)
280 not introduced 4%
(Japanese, Korean,
50 and Chinese)
0%

40 Web
Fraction of tweets, n_chars [136,140]
12% Fraction of tweets, n_chars [276,280]
30 the switch
8%

Faction
7

17

19
1

1
20

20

20

20
Jan

t
Jan

t
Oc

Oc
4%

Figure 8: Differences in differences setup for estimation of 0%

the effect of the introduction of 280 character limit on tweet Automated sources and third-party applications
Fraction of tweets, n_chars [136,140]
lengths. For the pre- (left) and post-switch (right) periods, 12% Fraction of tweets, n_chars [276,280]
the switch
we monitor the daily average tweet length, in the languages 8%

Faction
where the 280 limit is introduced (in orange), and in the lan-
4%
guage where it is not introduced (in blue). Dashed lines mark
the averages of the daily average tweet length in the four 0%

2017-1-1
2017-2-1
2017-3-1
2017-4-1
2017-5-1
2017-6-1
2017-7-1
2017-8-1
2017-9-1
2017-10-1
2017-11-1
2017-12-1
2018-1-1
2018-2-1
2018-3-1
2018-4-1
2018-5-1
2018-6-1
2018-7-1
2018-8-1
2018-9-1
2018-10-1
2018-11-1
2018-12-1
2019-1-1
2019-2-1
2019-3-1
2019-4-1
2019-5-1
2019-6-1
2019-7-1
2019-8-1
2019-9-1
2019-10-1
conditions.
date

Differences in differences estimation of the effect of
switch on tweets posted from different devices. To under-
stand the impact of the switch on tweets posted from differ-
ent devices, we fit a slightly different model
y ∼ treated ∗ period ∗ source,
where the source is a categorical variable representing
mobile devices, web devices, or automated sources and
third-party applications. By analogy to Equation 1, we
then isolate the total effect of switch on tweet length
posted by specific sources as the sum of the baseline
effect of the switch and the source-specific switch ef-
fect, treated : period + treated : period : source. Fit-
ting the model from Equation 2, we consistently measure a
largest increase in tweet length of 17.46% (95% CI [16.54%,
18.39%]) for web, followed by 9.76% (95% CI [8.88%,
10.66%]) for automated sources and third-party applica-
tions, and the smallest of 5.64% (95% CI [5.19%, 6.08%])
for mobile.
Next, we further investigate tweets posted by automated
sources and third-party applications. This is content likely
generated by bots or otherwise automated applications, as
opposed to the content generated by regular users. We show
for five biggest languages where 280 was introduced (En-
glish, Portuguese, Spanish, Arabic, and Indonesian), in Fig-
ure 10, tweet length histograms of tweets posted by third-
party applications and automated sources. The probability
is normalized relative to the baseline, the probability of
the same character length in the same language for regu-
lar sources (web and mobile). For example, +200% means
that a tweet of the given length is two times more likely to
be observed among the tweets posted by automated sources,
compared to the regular users. Again, we compare Jan-Oct
2017 and 2019 as pre- and post-periods. The patterns look
similar across languages. Automated sources and third-party
applications tweet longer tweets before the switch, and post
tweets in longer length ranges after the switch. The infliction
point at around 70 characters remains. Automated sources
Figure 9: Daily fraction (indicated with a circle), and 10-
day rolling average (solid line) of faction of tweets that have
between 276 and 280 characters (blue), and between 136 and
140 characters (red). The evolution is shown separately for
(2) Mobile applications, Web interface, and Automated sources
and third-party applications.
and third-party applications have a lingering peak at 140 af-
ter the switch in English, Arabic, and Indonesian.
4.4 RQ4: What are the syntactic and semantic
characteristics of long vs. short tweets?
Lastly, we aim to provide deeper insights into the nature of
long tweets tweeted after the switch. Why do users tweet
long tweets? What are their signature characteristics? To an-
swer those questions, we examine the content of the tweets.
We study tweets in English that were posted from mobile
and web devices during the two previously introduced pre-
(75.56M tweets) and post-switch (65.29M tweets) periods.
We annotated the tweets with LIWC (Pennebaker, Booth,
and Francis 2007) syntactic features (linguistic categories)
and semantic features (psychological, biological, and social
categories).
First, to characterize syntactic features of tweets, for all
tweets with a given number of characters, we measure the
occurrence frequency of different syntactic features–part of
speech (POS) tags among tweets of that length. In Fig-
ure 11, across all possible tweet lengths in the period be-
fore the switch ([1 − 140] characters), we observe the frac-
tion of tweets in that length that have at least one POS tag.
The dashed black line represents this quantity across tweets
posted in the period before the switch (under 140 charac-
ter limit), and the solid colored line represents this quantity
among tweets posted in the period after the switch (under
280 character limit).
 Tweet legth probability for automated sources and third-party applications Adverbs Article Assent AuxVb Conj Filler Negate
0.75 0.75 0.4
0.6 0.10 0.75 0.10
0.50 0.50
Jan-Oct 2017 Jan-Oct 2019 0.4 0.50 0.2
+300% +300% 0.05
+250% +250% 0.2 0.25 0.05 0.25 0.25
+200% +200%
+150% +150% 0.0 0.00 0.00 0.00 0.00 0.0
+100% +100%
English

+50% +50% 1 140 1 140 1 140 1 140 1 140 1 140 1 140

mobile and web mobile and web
-50% -50% Nonflu Numbers Prep Pronoun Quant Swear Verbs
-100% -100% 1.0
-150% -150% 0.03 0.4 0.100
-200% -200% 0.10
1 70 140 210 280 1 70 140 210 280 0.5 0.075
0.02 0.5 0.5
+300% +300% 0.05 0.2
+250% +250% 0.050
+200% +200% 0.01 0.025
+150% +150% 0.00 0.0 0.0 0.0 0.0
Portuguese

+100% +100% 1 140 1 140 1 140 1 140 1 140 1 140 1 140

+50% +50%
mobile and web mobile and web
-50% -50%
-100% -100%
-150% -150% Figure 11: Occurrence frequency of POS tags across tweets
-200% -200%
1 70 140 210 280 1 70 140 210 280 in different character lengths possible in the period before
+300% +300% switch ([1 − 140] characters). The dashed black line repre-
+250% +250%
+200% +200%
+150% +150% sents this quantity across tweets posted in the period before
+100% +100%
Spanish

+50% +50% the switch (under 140 character limit), and the solid colored
mobile and web mobile and web
-50% -50% line in the period after switch (under 280 character limit).
-100% -100%
-150% -150%
-200% -200% Adverbs Article Assent AuxVb Conj Filler Negate
1 70 140 210 280 1 70 140 210 280 1.0 0.6
0.75 0.75 0.75 0.15
+300% +300% 0.10
+250% +250% 0.50 0.50 0.50 0.10 0.4
+200% +200% 0.5
+150% +150% 0.25 0.05 0.05 0.2
+100% +100% 0.25 0.25
Arabic

+50% +50%
mobile and web mobile and web 0.00 0.00 0.0 0.00 0.00 0.0
-50% -50% 1 280 1 280 1 280 1 280 1 280 1 280 1 280
-100% -100%
-150% -150% Nonflu Numbers Prep Pronoun Quant Swear Verbs
-200% -200% 1.0 1.00 1.0
1 70 140 210 280 1 70 140 210 280 0.6
0.2 0.75 0.10
+300% +300% 0.04
+250% +250% 0.5 0.50 0.4 0.5
+200% +200% 0.1
+150% +150% 0.02 0.2 0.05
Indonesian

+100% +100% 0.25

+50%
mobile and web mobile and web
-50%
-100%
-150%
-200%
1 70 140 210 280
Number of characters
Figure 10: Tweet length probability for tweets posted by au-
tomated sources and third-party applications, before the 280
limit was introduced (on the left), and after (on the right).
The probability is normalized relative to the baseline, the
probability of same length in the same language for regular
sources: web and mobile.
Comparing the solid colored and dashed black lines
across different POS tags lets us isolate the effect of the
length limit on the content of the tweets. The largest gap
is observed for swear words and spoken categories (non-
fluencies, fillers, and assent), adverbs, and conjunctions–
nonessential parts of speech that are deleted in the process of
“squeezing in” a message to fit a length limit. No gap is ob-
served for verbs and negations, essential parts of speech that
are known to be disproportionately preserved in the shorten-
ing process (Gligorić, Anderson, and West 2019).
Figure 12 represents the same quantities after the switch
(under 280 character limit) across all possible tweet lengths
in this period ([1 − 280] characters). We note that there is
no counterfactual observation, i.e., we do not know what
the probability of observing a POS tag among 280 character
long tweets would be if the 280 character length limit was
lifted. Nonetheless, among the 280 character long tweets af-
ter the switch, we do observe patterns similar to those asso-
ciated with the 140 character long tweets before the switch.
+50% 0.0 0.0 0.0 0.0
1 280 1 280 1 280 1 280 1 280 1 280 1 280
-50%
-100%
-150%
-200%
1 70 140 210 280 Figure 12: Occurrence frequency of POS tags across tweets
posted in the period after the switch (under 280 character
limit) in different allowed character lengths ([1 − 280] char-
acters). POS tags are sorted alphabetically. Note the differ-
ent y-axis scales.
We observe a “dip” in the frequency of the spoken cate-
gories, conjunctions, and numbers among the 280 character-
long tweets, traces typical of “optimizing” a message to fit a
length limit.
In Figures 13 and 14 we measure the same quantities
for fine-grained subtypes of personal pronouns. This sug-
gests that personal pronouns I and you were most affected
by the 140 character limit (i.e., they were most likely to be
omitted). We observe a similar non-monotonic distribution
around 280 characters in Figure 14.
To summarize, 280 character tweets are syntactically sim-
ilar to 140 tweets before the switch: their usage is associated
with patterns that are indicative of “squeezing in” a message.
This is evidence indicating that they are generated by similar
writing processes as 140 character tweets were.
Next, in Figure 15 we study topics of tweets, measured by
LIWC categories describing psychological, biological, and
social categories. Across the studied topics, for each allowed
character length in the pre-switch and post-switch periods,
we measure the factor by which the topic is more frequent
in a given character length, compared to the overall topic
frequency. Categories are sorted by the value of this factor
 I SheHe They We You 4.0
Work
4.0
Money
0.2 Money Death

tweets before the switch, compared to among all tweets

0.15

tweets after the switch, compared to among all tweets

0.15 0.3 3.5 Achiev 3.5 Work

Factor by which the topic is more frequent among

Factor by which the topic is more frequent among

0.4 Death
Relig
Relig
Home
0.10 0.10 0.2 3.0 Home 3.0 Health
0.2 0.1 Humans Achiev
0.05 0.05 0.1 2.5 Health
Leisure 2.5 Humans
Family
Friends Friends
0.0 0.0 0.00 0.00 0.0 2.0 Family
Ingest
2.0 Leisure
Body
1 140 1 140 1 140 1 140 1 140 Body Ingest
1.5 Sexual 1.5 Sexual

Figure 13: Occurrence frequency of personal pronouns 1.0 1.0

across tweets in different character lengths possible in the 0.5 0.5

period before switch ([1 − 140] characters). The dashed
0.0 0.0
black line represents this quantity across tweets posted in 1 140 280 1 140 280
Number of characters Number of characters
the period before the switch (under 140 character limit), and
the solid colored line in the period after switch (under 280
character limit). Figure 15: For each allowed number of characters before the
I SheHe They We You switch (left), and after the switch (right), across 14 topics
0.3 0.3 0.4 measured with LIWC categories, we monitor the factor by
0.4 0.2 0.2 which the topic is more frequent among tweets with that
0.2
0.2 number of characters, compared to overall frequency. Cate-
0.2 0.1 0.1 0.1
gories are sorted by the value of this factor at 140 characters
0.0 0.0 0.0 0.0 0.0 (left), and 280 characters (right).
1 280 1 280 1 280 1 280 1 280

Figure 14: Occurrence frequency of personal pronouns
across tweets posted in the period after the switch (under
280 character limit) in different allowed character lengths
([1 − 280] characters). Personal pronouns are sorted alpha-
betically. Note the different y-axis scales.
at 140 characters before the switch, and at 280 characters
after the switch.
We note the personal concerns categories that were rel-
atively most frequent at 140 characters before the switch:
Work, Money, Achievement, Death, and Religion. The least
frequent topics at 140 characters, on the other hand, were
Biological categories: Sexual, Body, Ingestion, followed by
Family, Friends, and Leisure. An apparent association with
the importance of the message emerges: while tweets about
topics related to ordinary, overall more prevalent every-day
experiences use the longer tweets the least frequently, top-
ics related to more serious personal concerns use them the
most. Similarly to within-language, there is a within-topic
correlation between usage of 140 character length before the
switch, and subsequent usage of 280 character length after
the switch, with the ranking of the topics usage at the bound-
ary length only slightly changed (Spearman rank correlation
between topics 0.91, p = 7.30 ∗ 10−6 ), implying that 280
character tweets are also semantically similar to 140 tweets
before the switch.
5 Discussion and conclusions
To summarize, immediately after the switch we observe a
sharp decline in the frequency of 140 characters, and an in-
crease in the frequency of tweets long 280 characters (Fig-
ure 1). As 140 characters were becoming less prevalent af-
ter the switch, the prevalence of 280 was increasing af-
ter its introduction over a period of around 6 months (Fig-
ures 3 and 4). Taking advantage of the emergence of differ-
ing length restriction among languages, we estimate a signif-
icant increase in tweet lengths in languages where the switch
happened, compared to the control languages (Figure 8).
We note, however, that tweet lengths increased slightly in
the control languages as well. This is likely impacted by the
nature of how tweet length is counted at the character level
(as described in the Data Section), allowing mixed-character
tweets to be longer than 140 characters.
We observe that the more 140 was used in a language be-
fore the switch, the more 280 is used after (Figure 6). Disag-
gregation across languages reveals interesting temporal pat-
terns (Figure 7): In most languages, the prevalence of long
tweets seems to have settled, and is even decreasing again,
i.e., the peak of usage, or the “honeymoon phase” is over.
This is indicating the presence of a period of high usage
rates of the new feature, followed by a drop and saturation
to a constant level.
Different adoption patterns are observed between web and
mobile devices (Figure 9). Longer tweets are particularly
used on web clients. In the web interface, 140 was quickly
surpassed by 280, reaching around 4% at the end of the stud-
ied period. While slower adoption rates on mobile devices
could conceivably be linked to clients that did not update,
the fact that tweets were longer on web interface before the
switch indicates that there is simply a tendency for shorter
text on mobile phones. Automated sources and third-party
applications are the slowest to adapt. They posted longer
tweets before the switch, and post tweets in longer length
ranges after the switch (Figure 10). It is interesting to note
that automated sources and third-party applications write
longer than humans, but the effect is weaker at the boundary
(just under 140 before and just under 280 after the switch),
probably due to the “squeezing” of originally longer tweets
that humans do.
 Tweets long 280 characters are syntactically similar to
140 tweets before the switch: their usage is associated with
patterns that are indicative of “squeezing in” a message (Fig-
ures 11 and 12). This is evidence indicating that tweets close
to the new boundary are generated by similar writing pro-
cesses as 140 character tweets were before the switch. Sim-
ilarly to within-language, there is a within-topic correlation
between usage of 140 character length before the switch, and
subsequent usage of 280 character length after the switch
(Figure 15). Tweets 280 characters long are semantically
similar to 140 tweets before the switch: their usage is as-
sociated with the same topics. We note that this holds across
the range of long character lengths, and not only for 280
character long tweets specifically.
Given these findings, our guiding question—Is 280 the
new 140?—calls for a nuanced answer. On the one hand,
the emergence of the peak in the prevalence of 280 char-
acters, the fact that the languages that came close to the
140-limit also tend to come closer to the 280-limit, and the
traces of distinctive writing processes when “squeezing” a
message, absent in automated sources and third-party appli-
cations (Figure 10) all resonate with a narrative stating that,
yes, 280 is the new 140. On the other hand, the prevalence
of 280 is much less drastic than that of 140 used to be (Perez
2017)—only 4% for 280 characters after the switch, com-
pared to over 12% for 140 characters before the switch (for
tweets posted by Web clients). In this sense, while 280 may
indeed be considered the new 140, it is at the same time less
noticeable: in a nutshell, 280 is a less intrusive 140.
Finally, this evidence suggests that, just as the old 140-
character limit (Gligorić, Anderson, and West 2018), the
new 280-character limit impacts the writing style and con-
tent of tweets (Sen et al. 2019). The length constraint and
the resulting tweet-length distribution remain an important
dimension to consider in studies using Twitter data, as after
the switch the number of characters remains an important
variable, correlated with important properties of tweets in-
cluding topics, language, device, and the likelihood of being
an automated source of tweets.
Limitations. This study suffers from limitations that the 1%
stream is known to be susceptible to (Wu, Rizoiu, and Xie
2020), as certain accounts might be over-represented due to
the intentional or unintentional tampering with the Sample
API (Pfeffer, Mayer, and Morstatter 2018; Morstatter et al.
2013).
In our study, we use automated sources and third-party
applications as a proxy for bots. However, bot detection can
be more reliable using more sophisticated methods detect-
ing bots that use regular applications (Chavoshi, Hamooni,
and Mueen 2016; Kudugunta and Ferrara 2018; Yang et al.
2020). Syntactic and semantic analysis of tweets is limited
to tweets in English only, due to the lack of available tools
to support annotation. Future studies should measure these
characteristics in other languages, using language-specific
tools, or machine translation. Lastly, we note that in this
study, we study Twitter as a platform (tweets are sampled at
the community level), as opposed to users, whose timelines
are incomplete.
Future work. Future work should provide a better under-
standing of what user features are associated with adoption
(e.g., users’ age, number of followers, levels of activity). Are
the users who used 140 the same ones who are more likely
to use 280? Answering this question requires collection of
data beyond 1% sample, that contains complete records of
users’ tweets. However, here we caution against naive com-
parisons, as careful quasi-experimental designs are neces-
sary to truly isolate the effect of age. User age is corre-
lated with other factors–users who stay longer on the plat-
form might be in other ways fundamentally different from
younger users who joined more recently (i.e., there is “sur-
vivor bias” (Elton, Gruber, and Blake 1996)). Finally, our
study should be replicated several years from today, when
even more time has passed since the switch.
References
Artzi, Y.; Pantel, P.; and Gamon, M. 2012. Predicting responses
to microblog posts. In Proc. Conference of the North American
Chapter of the Association for Computational Linguistics.
Bakshy, E.; Hofman, J. M.; Mason, W. A.; and Watts, D. J. 2011.
Everyone’s an influencer: Quantifying influence on twitter. In Proc.
ACM International Conference on Web Search and Data Mining.
Berger, J., and Milkman, K. L. 2012. What makes online content
viral? Journal of Marketing Research 49(2):192–205.
Boot, A. B.; Sang, E. T. K.; Dijkstra, K.; and Zwaan, R. A. 2019.
How character limit affects language usage in tweets. Palgrave
Communications 5(1):1–13.
Boyd, D.; Golder, S.; and Lotan, G. 2010. Tweet, tweet, retweet:
Conversational aspects of retweeting on twitter. In 2010 43rd
Hawaii international conference on system sciences, 1–10. IEEE.
Centola, D.; Becker, J.; Brackbill, D.; and Baronchelli, A. 2018.
Experimental evidence for tipping points in social convention. Sci-
ence 360(6393):1116–1119.
Chang, H.-C. 2010. A new perspective on twitter hashtag use: Dif-
fusion of innovation theory. Proceedings of the American Society
for Information Science and Technology 47(1):1–4.
Chavoshi, N.; Hamooni, H.; and Mueen, A. 2016. Debot: Twitter
bot detection via warped correlation. In Icdm, 817–822.
Ciot, M.; Sonderegger, M.; and Ruths, D. 2013. Gender inference
of twitter users in non-english contexts. In Proceedings of the 2013
Conference on Empirical Methods in Natural Language Process-
ing, 1136–1145.
Clarke, I., and Grieve, J. 2019. Stylistic variation on the donald
trump twitter account: A linguistic analysis of tweets posted be-
tween 2009 and 2018. PloS one 14(9):e0222062.
Cunha, E.; Magno, G.; Comarela, G.; Almeida, V.; Goncalves,
M. A.; and Benevenuto, F. 2011. Analyzing the dynamic evolution
of hashtags on twitter: a language-based approach. In Proceedings
of the workshop on language in social media (LSM 2011), 58–65.
Doyle, G.; Yurovsky, D.; and Frank, M. C. 2016. A robust frame-
work for estimating linguistic alignment in twitter conversations.
In Proceedings of the 25th international conference on world wide
web, 637–648.
Eisenstein, J. 2013. What to do about bad language on the in-
ternet. In Proc. Conference of the North American Chapter of the
Association for Computational Linguistics.
Elton, E. J.; Gruber, M. J.; and Blake, C. R. 1996. Survivor
bias and mutual fund performance. The review of financial stud-
ies 9(4):1097–1120.
 Flekova, L.; Preoţiuc-Pietro, D.; and Ungar, L. 2016. Exploring
stylistic variation with age and income on twitter. In Proceedings
of the 54th Annual Meeting of the Association for Computational
Linguistics (Volume 2: Short Papers), 313–319.
Garimella, K.; Weber, I.; and De Choudhury, M. 2016. Quote
rts on twitter: usage of the new feature for political discourse. In
Proceedings of the 8th ACM Conference on Web Science, 200–204.
Gligorić, K.; Anderson, A.; and West, R. 2018. How constraints
affect content: The case of twitter’s switch from 140 to 280 charac-
ters. In Proc. International Conference on Web and Social Media.
Gligorić, K.; Anderson, A.; and West, R. 2019. Causal effects of
brevity on style and success in social media. Proc. ACM Hum.-
Comput. Interact. 3(CSCW).
Guerini, M.; Strapparava, C.; and Özbal, G. 2011. Exploring text
virality in social networks. In Proc. International Conference on
Web and Social Media.
Hu, Y.; Talamadupula, K.; and Kambhampati, S. 2013. Dude,
srsly?: The surprisingly formal nature of twitter’s language. In
Proc. International Conference on Web and Social Media.
Jaidka, K.; Zhou, A.; and Lelkes, Y. 2019. Brevity is the soul of
twitter: The constraint affordance and political discussion. Journal
of Communication 69(4):345–372.
Jin, H., and Liu, H. 2017. How will text size influence the length
of its linguistic constituents? Poznan Studies in Contemporary Lin-
guistics 53(2):197–225.
Kooti, F.; Mason, W. A.; Gummadi, K. P.; and Cha, M. 2012a. Pre-
dicting emerging social conventions in online social networks. In
Proceedings of the 21st ACM international conference on Informa-
tion and knowledge management, 445–454.
Kooti, F.; Yang, H.; Cha, M.; Gummadi, P. K.; and Mason, W. A.
2012b. The emergence of conventions in online social networks.
In ICWSM.
Kudugunta, S., and Ferrara, E. 2018. Deep neural networks for bot
detection. Information Sciences 467:312–322.
Laib, N. 1990. Conciseness and amplification. College Composi-
tion and Communication 41(4):443–459.
Lamprinidis, S.; Hardt, D.; and Hovy, D. 2018. Predicting news
headline popularity with syntactic and semantic knowledge using
multi-task learning. In Proc. Conference on Empirical Methods in
Natural Language Processing.
Levinson, P. 2011. The long story about the short medium: Twit-
ter as a communication medium in historical, present, and future
context. Journal of Communication Research 48:7–28.
McLuhan, M. 1964. The extensions of man. New York.
Morstatter, F.; Pfeffer, J.; Liu, H.; and Carley, K. M. 2013. Is the
sample good enough? comparing data from twitter’s streaming api
with twitter’s firehose. arXiv preprint arXiv:1306.5204.
Murthy, D. 2012. Towards a sociological understanding of social
media: Theorizing twitter. Sociology 46(6):1059–1073.
Page, R. 2012. The linguistics of self-branding and micro-celebrity
in twitter: The role of hashtags. Discourse & communication
6(2):181–201.
Pancer, E., and Poole, M. 2016. The popularity and virality of
political social media: hashtags, mentions, and links predict likes
and retweets of 2016 u.s. presidential nominees tweets. Social In-
fluence 11(4):259–270.
Pavalanathan, U., and Eisenstein, J. 2016. More emojis, less:) the
competition for paralinguistic function in microblog writing. First
Monday.
View publication stats
Pennebaker, J. W.; Booth, R. J.; and Francis, M. E. 2007.
Liwc2007: Linguistic inquiry and word count. Austin, Texas: liwc.
net.
Perez, S. 2017. Twitter’s doubling of character count from 140 to
280 had little impact on length of tweets. https://cutt.ly/gfoIaY1.
Perrin, A., and Anderson, M. 2019. Share of us adults using social
media, including facebook, is mostly unchanged since 2018. Pew
Research Center 10.
Pfeffer, J.; Mayer, K.; and Morstatter, F. 2018. Tampering with
twitter’s sample api. EPJ Data Science 7(1):50.
Rimjhim, and Chakraborty, R. 2018. Characterizing user reactions
towards twitter’s 280 character limit. In Proceedings of the 10th
Annual Meeting of the Forum for Information Retrieval Evaluation,
FIRE’18, 48–51. ACM.
Rosen, A., and Ihara, I. 2017. Giving you more characters to
express yourself. https://blog.twitter.com/official/en_us/topics/
product/2017/Giving-you-more-characters-to-express-y\ourself.
html.
Rosen, A. 2017a. Tweeting made easier. https://goo.gl/DYzBji.
Rosen, A. 2017b. Tweeting made easier. https://blog.twitter.com/
official/en_us/topics/product/2017/tweetingmadeeasier.html.
Sen, I.; Floeck, F.; Weller, K.; Weiss, B.; and Wagner, C. 2019. A
total error framework for digital traces of humans. arXiv preprint
arXiv:1907.08228.
Shapp, A. 2014. Variation in the use of twitter hashtags. New York
University 1–44.
Sloane, B. S. 2003. Say it straight: Teaching conciseness. Teaching
English in the Two Year College.
Sylwester, K., and Purver, M. 2015. Twitter language use re-
flects psychological differences between democrats and republi-
cans. PloS one 10(9):e0137422.
Tan, C.; Lee, L.; and Pang, B. 2014. The effect of wording on
message propagation: Topic- and author-controlled natural experi-
ments on twitter. In Proc. Annual Meeting of the Association for
Computational Linguistics.
Vardi, A. D. 2000. Brevity, conciseness, and compression in ro-
man poetic criticism and the text of gellius’ noctes atticae 19.9. 10.
American Journal of Philology.
Wang, S. A., and Greenwood, B. N. 2020. Does length impact
engagement? length limits of posts and microblogging behavior.
Length Limits of Posts and Microblogging Behavior (February 12,
2020).
Wasike, B. S. 2013. Framing news in 140 characters: How so-
cial media editors frame the news and interact with audiences via
twitter. Global Media Journal: Canadian Edition 6(1).
Wikström, P. 2014. # srynotfunny: Communicative functions of
hashtags on twitter. SKY Journal of Linguistics 27.
Wu, S.; Rizoiu, M.-A.; and Xie, L. 2020. Variation across scales:
Measurement fidelity under twitter data sampling. In Proceedings
of the International AAAI Conference on Web and Social Media,
volume 14, 715–725.
Yang, K.-C.; Varol, O.; Hui, P.-M.; and Menczer, F. 2020. Scal-
able and generalizable social bot detection through data selection.
In Proceedings of the AAAI Conference on Artificial Intelligence,
volume 34, 1096–1103.
Zhou, A., and Xu, S. 2019. Remaking dialogic principles for the
digital age: The role of affordances in dialogue and engagement.
SocArXiv.