Does a full stop make a text sound angry?

Someone replied “Sure.” and you have been looking at the dot for a while. I run Subtext, which reads messages for tone, so people send me this question more than any other, usually with a screenshot attached.

In a very short message a full stop can make the reply read as slightly less sincere, and the effect is real but small. It has never been shown to mean anger, whatever the are they mad at me spiral says. In the one study that varied length, it was no longer significant once the message was long enough that punctuation looks normal there. Almost everything else you have read about this comes from one study on 126 university students that measured something narrower than the headlines suggested.

Americans call it a period. I will call it a full stop from here.

What the original study actually did

Gunraj and colleagues at Binghamton University built 16 exchanges where one person made a statement and then asked a question, along the lines of a friend saying they have spare tickets and asking if you want to come¹. The reply was one of four words: okay, sure, yeah or yup. Half the time it ended in a full stop and half the time it ended in nothing.

Participants rated the responder’s sincerity on a seven-point scale. Full stop versions averaged 3.85. Unpunctuated versions averaged 4.06. That is a difference of 0.21 of a point, on a scale of one to seven, in a one-word reply.

What I find clever about the design is the bit that usually gets dropped. Half the participants saw the exchanges as text messages on pictures of phones, and half saw the same exchanges as handwritten notes on loose-leaf paper. The full stop produced no significant sincerity difference in the handwritten condition. So the dot does not look harsh in itself. Readers were applying a convention that belongs specifically to text messaging.

What the study did not do, in a list, because the gap between this and the coverage is what this whole piece is about:

• It did not measure anger, rudeness, hostility or passive aggression. It measured sincerity.
• It did not test messages longer than one word.
• It did not compare age groups. Everyone was a Binghamton undergraduate.
• It did not test any language other than English.
• It did not tell us anything about what the sender meant. The sender was fictional.

I’d flag one more thing. The widely repeated claim that an exclamation mark makes a reply seem more sincere was described in the university’s press release as unpublished follow-up work². Several outlets reported it as an established finding. It has not been published as one. This is the exact gap Subtext is built to sit inside. The mark measures something narrow and specific, and it is rarely the thing everyone assumes.

The extension that made it more credible

Houghton, Upadhyay and Klin, from the same lab, ran longer and more naturalistic exchanges, and varied the content of the reply itself so that some were negative, like “nope”, and some ambiguous, like “maybe”³. Terminal full stops pushed interpretations in a more negative and less enthusiastic direction, and made negative replies read as more abrupt.

This is where abruptness and negative reading gain actual experimental support, which is why it matters to me. It is also not an independent replication, since two of the original authors are on it, which makes it replication plus extension within the same lab. Abruptness is closer to what Subtext looks for than insincerity is, since abruptness is about how a reply lands and not what the sender secretly felt.

The study that explains the whole thing

The paper I found most useful in this literature came out in 2025, from Kemp, Kovacic and Beyersmann in Australia, and it is the one nobody has heard of⁴.

They showed 200 Australian undergraduates 30 message exchanges, varying valence, whether the final message ended in a full stop, and message length: one word, three to four words, or six to eight words. Participants rated how they thought the sender felt.

Messages with a full stop were rated more negatively than messages without. Short and medium messages were rated more negatively than long ones. And for the six to eight word messages, the difference the full stop made was not statistically significant.

The authors are refreshingly candid about the size of what they found. The negative effect for short and medium messages, in their words, was “numerically relatively small”, running between 0.11 and 0.37 of a point on a seven-point scale.

The finding, in the researchers’ own words “Numerically relatively small.”

How Kemp, Kovacic and Beyersmann describe the negative effect of a full stop on short and medium messages: between 0.11 and 0.37 of a point on a seven-point scale.

This is the most useful evidence in the field because it tests an explanation, where the other studies mostly repeat a demonstration, and not because it is the biggest. If a full stop were a symbol meaning “I am annoyed”, it should still mean that at eight words. Instead the effect was no longer significant at the longest length they tested, which is about where a full stop stops being optional. It is one study with students, so treat six to eight words as what they tested, not a safe word count. Subtext runs on the same logic. Whether a mark is doing any work at all depends on everything around it, not on the mark itself.

Why an optional full stop carries meaning

Subtext reads a full stop against the length of the message, what the message says and who it is going to, because the same dot in “Sure.” and in “I’ll be there at seven.” is not doing the same work. That is possible only because of what “optional” means here. The message bubble has already ended your sentence for you. When you type “Sure” and hit send, the interface makes the endpoint obvious, so the dot is doing no grammatical work. Ling and Baron’s corpus work found roughly 29 per cent of message-final sentences ended in a full stop, against about 54 per cent of sentences in the middle of a multi-sentence message⁵. Ending the message is itself the punctuation.

Once something is optional, readers assume its presence was a choice. The reasoning goes that they did not need that mark and put it there anyway, so it might be doing something.

That something need not be anger. It can be closure, seriousness, restraint, formality or distance. Which one wins depends entirely on the words around it and on who is sending it.

This is the part generic software gets wrong when it treats punctuation as a dictionary. A tool that flags every terminal full stop is counting characters and not reading tone, which is why most tone checkers never check your tone.

The best evidence that the dot is not hostile

Albritton asked 41 undergraduates to write text messages in response to four scenarios: a friend getting a job, a friend being rejected for a job, ambiguous job news, and a group project request⁶. Then he counted what they produced.

Message-final full stops turned up most often in the consoling scenario. Eleven of them, at 0.32 per message, against three in the happy scenario and one in the ambiguous one. People reached for the full stop when they were being serious with someone who was upset.

That does not overturn the Binghamton finding, because the two studies ask different questions. Albritton measured what writers chose to type. Gunraj measured what readers inferred. A writer can intend gravity while a reader hears flatness, the same gap behind does this text sound rude, which is roughly the entire problem with texting, and the gap those two studies describe is the one Subtext exists to close.

What it does kill is the literal claim that the mark is intrinsically hostile. Albritton’s own conclusion is that there is no obvious emotional meaning attached to the full stop, and his data shows it turning up in congratulation and consolation.

Is this a generational thing?

Probably less than you think, and the research does not currently support the confident version.

The original experiment had no older comparison group, so it cannot tell you that young people react differently from old people. It can only tell you that some young adults were sensitive to the cue. The central logical error in most of the coverage is treating “young people in an experiment showed X” as if it got you to “young people show X and older people do not”. You would need participants spanning cohorts, identical materials, and a statistical interaction between age and punctuation. I couldn’t find a large peer-reviewed experiment that did that.

The nearest thing to an age check is a 2025 Yale undergraduate thesis, which found no overall sincerity penalty for its lowercase-plus-full-stop construction, including among participants aged 30 and under⁷. That is student work with a broader manipulation than the original, so it is a useful signal, not a verdict.

The account I find most convincing is about exposure rather than birth year. Spend years in group chats where nobody uses final punctuation and adding it becomes a marked choice. Age correlates with that experience, but so do platform, profession, first language and personal style. Plenty of people in their twenties punctuate everything and plenty of people in their sixties do not. This is why Subtext reads a message against the thread it sits in and not against a rule about age. The same full stop from someone who never uses one and from someone who punctuates every message is two different signals.

What the other marks are doing

Subtext reads every one of these marks the same way it reads the full stop, against the message it sits inside and not as a fixed code with one meaning.

Exclamation marks. Waseleski’s content analysis of 200 exclamations across two professional discussion lists concluded they rarely marked excitability and more often marked friendly interaction⁸. Albritton’s writers used them most in the congratulatory scenario, at one per message, against 0.19 per message in the task scenario⁶. The mark supplies emphasis and heightened involvement. Whether that reads as warm depends on the words, since a cheerful message and a sharp instruction can both carry one.

Ellipses. Raclaw’s work treats the three dots as a discourse-management device: incompletion, hesitation, an unfinished turn, holding the floor⁹. That is closer to trailing off than to hostility. In one small study of 40 Serbian students of English, ellipsis-final messages were rated lowest of four endings on satisfaction and confidence, at 3.02 against 4.14 for a full stop and 4.39 for no punctuation¹⁰. Small sample, second-language readers, so directional only. “Hesitant” and “leaving something unsaid” describe the mark better than “passive-aggressive” does, though in a conversation that is already tense, an unfinished signal is easy to read as ominous.

Repeated punctuation at work. This is the one with a real cost attached. Sidi, Glikson and Cheshin found that repeated question marks in a work email correlated with lower perceived competence at minus 0.36, and that competence, rather than warmth, mediated a drop in willingness to help the sender¹¹. They ran it twice, once with 142 Dutch undergraduates reading an English email and once with 94 Israeli MBA students reading real Hebrew work emails. In the second study the penalty appeared for clear messages and vanished for ambiguous ones. Note the contrast with the texting research. In email between strangers informality costs you, and in texts between friends formality does.

A full stop after every word. Poirier, Cook and Klin tested this separately and it is a different phenomenon from a single terminal dot¹². They compared three or four word messages written normally with the same words punctuated after each one. The heavily punctuated versions were rated substantially more frustrated or disgusted. Sending each word in its own bubble did something similar. Putting each word on its own line inside one bubble did not, which rules out the explanation that it is just about taking up more space. Their reading is that both devices cue dramatic pauses, so the reader hears segmented speech.

Capital letters. Heath’s work on caps treats them as marking prosodic prominence, roughly where speech would add stress or volume¹³. “That was SO good” is not read as screaming one word. Whole messages in capitals can cue shouting because sustained emphasis invites that reading, but that is an inference from context, not the literal meaning of uppercase.

Does this work the same way in other languages?

Subtext reads tone in 17+ languages, so the English full stop finding is not something it can assume travels unchanged. I couldn’t find a verified experiment that takes the original manipulation and reproduces it with terminal punctuation in Japanese, Korean, Chinese, Arabic or any non-Latin script. That absence should make everyone more careful than they are being.

What does exist is descriptive. Androutsopoulos and Busch analysed WhatsApp conversations among German secondary school students and found the message-final full stop taking on interactional work, contextualising termination, distance or an official register, alongside the formal norms taught at school¹⁴. Busch’s related paper treats punctuation as a resource for organising turns and closure¹⁵. One striking figure from that work is that German adolescents’ informal digital writing carried around 3.7 full stops per 1,000 words, against around 72.1 per 1,000 in their formal writing.

In Japan the phenomenon has a name. Maru-hara, or full-stop harassment, describes younger people reading a message-ending 。 from an older colleague as cold or intimidating, and it became a subject of national coverage in April 2024¹⁶. The evidence behind it is company surveys, not research, and one of them cuts against the story. A 2024 survey of 500 people aged 20 to 60 found 64.8 per cent had never really been bothered by a full stop from a senior colleague, with 9.0 per cent citing intimidation¹⁷. Younger respondents were more sensitive, but the majority in every age group was unbothered.

A full stop’s meaning depends on whether the mark is routinely omitted in casual messaging in that language, what other ending devices exist, and what the local community has decided those alternatives mean. An effect generated by the optionality of the English full stop has no automatic right to transfer.

How the coverage went wrong

I’ll name it, because you have probably read the wrong version.

The 2016 study measured sincerity. Time ran the framing that a full stop makes you sound angry. Quartz published a piece on why a full stop in a text makes you sound angry. That paper never measured anger. Kemp and colleagues catalogue the escalation directly in their 2025 paper. The full stop has been called weaponised in the New York Times, declared pissed in the New Republic, and blamed for making a texter a monster⁴.

The handwritten control, which is the single most informative thing in the original design, disappeared from most retellings. Without it, readers cannot see that the finding is about expectations attached to a medium.

And perception got blurred with intention throughout. The experiment shows how some readers interpreted a cue. It says nothing about whether people who type full stops are less sincere. The sender in the study did not exist.

So what do you do with the dot

If you are the reader, a full stop in a short reply is weak evidence of anything. It is worth roughly 0.2 of a point on a seven-point scale in a controlled experiment, and the person may simply have a keyboard that adds it. If you are building a case from punctuation, you are building it from the thinnest material available.

If you are the writer, the mark carries more weight in a one-word reply than in a full sentence, so “Sure.” is doing more work than you probably intend and “I’ll be there at seven.” is doing almost none. If you want warmth in a short reply, the words are a stronger instrument than the punctuation. Flat words in a short reply are what people usually mean by dry texting.

Subtext flags this when it matters and stays quiet when it does not, which is most of the time. The useful version of tone checking is knowing which signals are weak, and a terminal full stop is one of the weakest ones there is.

Try Subtext in your browserGet the appTry Subtext in your browser

The full stop is a cue, not a code, and the person who sent it probably was not thinking about it at all.

---

Think I have read a study wrong, or know research on this I have missed? Tell me on LinkedIn.

Samet Durgun is the co-founder of Subtext, an app that catches the emotional tone of your messages and rewrites them in your own voice. He’s based in Berlin.

Sources

Numbered in the order they appear above. Sample sizes and figures are given where they were verified against the primary source, and the text says so where they were not. Checked 28 August 2026.

1. Gunraj, D. N., Drumm-Hewitt, A. M., Dashow, E. M., Upadhyay, S. S. N., and Klin, C. M. (2016). Texting insincerely: The role of the period in text messaging. Computers in Human Behavior, 55, 1067 to 1075.
2. Binghamton University, State University of New York. Study: Text messages that end in a period seen as less sincere. News release, 8 December 2015. Source of the unpublished exclamation mark aside.
3. Houghton, K. J., Upadhyay, S. S. N., and Klin, C. M. (2018). Punctuation in text messages may convey abruptness. Period. Computers in Human Behavior, 80, 112 to 121.
4. Kemp, N., Kovacic, R., and Beyersmann, E. (2025). Is the period really “pissed”? The effect of punctuation and message length on perceptions in digital communication. Telematics and Informatics, 97, article 102241.
5. Baron, N. S., and Ling, R. (2011). Necessary Smileys and Useless Periods: Redefining Punctuation in Electronically-Mediated Communication. Visible Language, 45(1/2), 45 to 67.
6. Albritton, A. (2022). What’s the Point? A Study of Full-Stop Use in Text Messages in Varying Emotional Contexts. Online Journal of Communication and Media Technologies, 12(1), e202203.
7. O’Bryan, A. (2025). wow, that’s really interesting. Or, the Tonal Effect of Capitalization and Punctuation in Online Communication. Yale University, BA thesis. Student research, not peer reviewed.
8. Waseleski, C. (2006). Gender and the Use of Exclamation Points in Computer-Mediated Communication. Journal of Computer-Mediated Communication, 11(4), 1012 to 1024.
9. Raclaw, J. (2006). Punctuation as Social Action: The Ellipsis as a Discourse Marker in Computer-Mediated Communication. Proceedings of the Annual Meeting of the Berkeley Linguistics Society, 299 to 306.
10. Milisavljević, M. (2025). I’m Not Angry…: Perceptions of Punctuation Marks in Online Text Messages by Serbian Students of English. Belgrade BELLS, 17. Forty participants; the full stop versus no punctuation comparison was not statistically significant.
11. Sidi, Y., Glikson, E., and Cheshin, A. (2021). “Do You Get What I Mean?!? The Undesirable Outcomes of (Ab)Using Paralinguistic Cues in Computer-Mediated Communication.” Frontiers in Psychology, 12, article 658844.
12. Poirier, R. C., Cook, A. M., and Klin, C. M. (2025). Read. This. Slowly: mimicking spoken pauses in text messages. Frontiers in Psychology, 16, article 1410698.
13. Heath, M. (2018). Orthography in social media: Pragmatic and prosodic interpretations of caps lock. Proceedings of the Linguistic Society of America, 3, article 55.
14. Androutsopoulos, J., and Busch, F. (2021). Digital punctuation as an interactional resource: the message-final period among German adolescents. Linguistics and Education, 62, article 100871.
15. Busch, F. (2021). The interactional principle in digital punctuation. Discourse, Context and Media, 40, article 100481.
16. The Japan Times. Fear of a full stop? Young Japanese say drop the period in your texts. 2 April 2024. Media reporting, not research.
17. FORiT Inc. survey on maru-hara, covered by ITmedia NEWS, 8 May 2024. Web survey of 500 respondents nationwide. A marketing company survey, not peer-reviewed research.