The meanings we glean in conversation are often, maybe mostly, not found in the words spoken, but in how they’re said, and in the spaces between them. Tone of voice, and cadences created by shifts in speed, volume, and pitch, let listeners know whether “Nice job,” is complimentary or sarcastic, or whether “Wow” shows that you’re impressed or underwhelmed. The literal meaning of words is their message, and everything about how words are said is the metamessage. Metamessages communicate how you mean what you say.

More and more conversations are taking place on screens—via texting, Facebook, Instagram, Snapchat, Twitter, email, and myriad other platforms. Some of these written conversations make up for the lack of voicing with conventions that mimic speech, like exclamation points, CAPS, and repetition of words or letters. I can be “so happy!!!!!!!” or “sooooo happy” or “SO happy” or “sosososo happy” or even “SOSOSOSOOOOOO happy!!!!!!!!!!!!!!!” Emoticons, emojis, and gifs help, too. But these visual signals are only the tip of the metamessage iceberg.

Human beings are always in the business of making meaning and interpreting meaning. Because there are options to choose from when sending a message, like which platform to use and how to use it, we see meaning in the choice that was made. But because the technologies, and the conventions for using them, are so new and are changing so fast, even close friends and relatives have differing ideas about how they should be used. And because metamessages are implied rather than stated, they can be misinterpreted or missed entirely.

Different generations are particularly apt to perceive different metamessages in the same words or actions. For example, a Sri Lankan woman living in London told me of hosting her sister and her sister’s teenage daughter. When the girl refused to go out after 3:00 p.m., because that’s when her friends back home got active on WhatsApp, the sisters perceived the metamessage, “I’m not really here. My head and my heart are elsewhere.” But I suspect the girl’s perspective was more like: “I’m here with you—that’s a given—but I also want to stay connected to my friends, and I can’t bear the thought of being left out.”

I heard a great range of viewpoints on social media while interviewing over 80 women (ranging in age from 9 to 97) for a study of friendship. And my students at Georgetown University have helped me understand how they manage, and sometimes are tripped up by, the metamessages communicated over social media—and how habits and assumptions can differ.

For instance, when Kate Lucey’s sister had not given birth by her due date, she kept family and friends apprised of what was happening with her pregnancy by posting frequently on Facebook. In response, even distant relatives and casual friends posted well wishes and encouraging thoughts. Kate felt that these postings sent her sister a precious metamessage: Many people love her and care about her pregnancy. The posts strengthened her network of support. Kate was stunned to learn that her roommate perceived a negative metamessage in her sister’s use of Facebook. She said she would have been offended to be kept up to date about a pregnancy that way; she would think, “Geez, why didn’t you call?” Kate’s roommate reasoned that posting on Facebook is so easy that it means nothing. A phone conversation reflects and creates a meaningful relationship, whereas Facebook creates a false sense of intimacy, not a real relationship.

The impression that posting on Facebook is insincere because it’s too easy sounds a lot like the explanation a friend gave me (over email) for disliking “THAT DRAT ‘xoxo,’ which means nothing, just keys to hit.” This perspective–if something is too easy it’s meaningless or insincere—makes sense. Taking time sends a metamessage of caring. Yet I like ‘xoxo’ (or my personal variant, ‘xxoo’), and use it, though only with friends who use it first. I think xoxo fills a need: It’s a more affectionate way to close an email than “Best” but not as fervent as “Love,” which, in any case, some people don’t feel comfortable using with friends, though some do. Maybe it strikes some as fake because it’s a substitution, standing for “love and kisses” but not denoting it. Or maybe it seems too cutesy. Whatever the source of these impressions, deciding how to sign off a message becomes a challenge, since any choice you make will send metamessages that you may not intend or suspect.

I was shocked to realize that my students might be perceiving metamessages opposite from what I intend when I reply to their email requests and queries. One of them pointed out that when he emails professors, he begins with a salutation (“Dear Professor Smith”) and a greeting like “I hope you’re enjoying the weekend.” Only then does he explain his request—in detail—followed by a friendly closing, again about the weekend or the weather, before wrapping up with “Sincerely” or “Yours truly” above his name. (I recognized his description: That’s the way most of my students’ emails look.) The student then complained that many of his professors’ responses omit the salutation, greeting, full explanation, friendly closing, signature, and name. All he receives is a naked reply to his question, and usually a cryptic one at that. I realized that I do this, too–or did, before he, and others in class discussion, opened my eyes. I had assumed that dispensing with those formalities sends a metamessage of casual friendliness, more or less the way I use email with close friends and family. But students regard email as formal, so the omission of those niceties often strikes them as disrespectful, even rude. (Now I go back and add the trimmings before pressing SEND.)

It’s not surprising that professors’ and students’ practices would vary, given the difference in power and age. But even best friends can have very different ideas about appropriate ways to use social media. Noelle Miesfeld and Rachel Jacobson had been close friends since college, and they stayed in close touch after graduation, often having long telephone conversations, catching up. After a number of years, however, they began communicating more through texting. This meant more frequent conversations—often daily or even multiple times a day. So Noelle was surprised when Rachel registered a complaint: She’d been telling Noelle about a problem, and she felt that Noelle’s responses seemed too casual and brief to show real concern. Rachel missed her caring, emotionally supportive friend. They traced the trouble to their contrasting assumptions about texting. To Noelle, comforting closeness resided in the frequency of their exchanges. To Rachel, frequency didn’t substitute for the expression of feeling and detailed discussion of her situation that they’d shared in the past—the kinds of conversations that Noelle didn’t feel could or should take place through texting.

When deciding which platform to use and how to use it, as well as how to interpret communications you receive—or don’t receive—you have to know which platforms your friends tend to use and how they use them. Some will answer texts but not emails. Others don’t check their phones regularly, so you can’t rely on texting to reach them. The proliferation of platforms means more options to exploit but also more opportunities for your messages to be misinterpreted. How quickly does a particular friend usually reply to a text or email? What does the lack of an immediate response mean? So firm is some people’s expectation of a quick reply to a text that any lapse carries meaning. A student, telling me about a friend who, in her words, “stopped talking” to her, said, by way of explanation, “She’d text me back two days later.” To her, a two-day delay was tantamount to not talking at all.

Silence can be a deliberate communication. One woman said of a text she received from a friend, “I was so annoyed, I’m just not responding. I just didn’t answer.” Yet in other cases, silence doesn’t mean anything; it results from circumstances. A young woman thought her boyfriend’s delay in responding meant he was angry at her. It turned out his phone battery had run out.

Perceived metamessages of annoyance can snowball, whether or not they were intended. A student recounted in an interview a Facebook message exchange with a friend that, she felt, should have been straightforward but became complicated. She sent a message suggesting they go running later than planned. The friend messaged back, “I guess that’s okay. Fine, see you then.” That little opener, “I guess,” and the unenthusiastic “okay” seemed to imply that she wasn’t thrilled with the change of plans. So the student responded, “I’m sorry, I didn’t mean to be difficult. I’m more flexible than I made it sound. I can actually just do one o’clock like we planned.” That set off a back-and-forth that went four rounds—“No no no let’s do the time you wanna do,” “No no no let’s do the time we said.” Hair-tearing exchanges like these are particularly common among girls and women because, as the student pointed out, they often worry about coming across as too demanding and hurting others’ feelings.

My student Holly DiClemente explained how her peers make creative use of digital features to avoid hurting friends’ feelings. One example is ghost reading—reading a message without opening it, by just reading the preview in the text app, or on your lock screen. If a phone automatically sends “read receipt” notices to let others know their messages have been opened and, presumably, read, ghost reading comes in handy to manage the implied metamessages. If you see from the text notification that a friend is asking if you want to hang out, and you don’t want to but you don’t want to hurt her feelings, you can open the message later and tell her you’re sorry you just got it. The “read receipts” feature can also be used to show you’re mad; it lets someone know you’ve read their message and are intentionally not responding—a visual virtual snub. But there, too, they might think you’re mad when you’re not. Maybe you saw that the message was long, so you put off replying until you’d have more time to read it, or to write a thoughtful response.

Metamessages—intended or not—can reside in just about any aspect of digital communication, even something as minor and automatic as listing recipients’ names when sending email. If you enter the addresses on the “to” or “cc” line, everyone who receives it can see not only who else is getting the message but also what order you put the names in. Two women I interviewed together, Lucy McBride and Annie Finnell, pointed out some of the resulting metamessages: “You feel special when you’re the first one on there,” Lucy said. Annie agreed: “You do. And you feel not special when you’re the last one.” Lucy added, “Because if you’re dead last, it’s like they were thinking, ‘Who am I forgetting?’” To avoid offending those whose names come later or last, you can put recipients’ email addresses on the “bcc” line. But that sends a metamessage, too. It’s what people do when the list of recipients is very long, so “It looks like you’re inviting all of Northwest Washington!”

Anytime there are multiple recipients, metamessages can get complicated. I was part of a group that used email to confer about a joint project. In one such exchange, a member of our group expressed well wishes to another about a medical condition, after which all the others chimed in, echoing the expressions of concern. I was puzzled. I didn’t know whether this flurry of well wishes was related to the ongoing medical challenge I knew this group member had been experiencing, or whether there had been a new and dangerous development I didn’t know about. I asked another group member, who said she was not aware of any new developments either, but wanted to express her general support for our ailing friend. I went ahead and sent my own well wishes, cc’ing only the one who first expressed concern. But then I worried that the others would think me callous and uncaring. I rather regretted having foregone the option of allowing—or forcing--everyone to overhear my well wishes.

Every word we speak is chock-full of metamessages telling others not only how we mean what we say but also who we are. It has always been that way; it’s the only way language can work to communicate ideas and negotiate relationships. With social media, we have ever more ways to do this—and ever more things to worry about, to make sure the self we’re displaying is the one we believe we are—or the one we want to be.

This article has been partially adapted from Deborah Tannen’s forthcoming book You’re The Only One I Can Tell: Inside the Language of Women’s Friendships.