A new Facebook research study shows that 70 percent of us have written a comment, or Facebook status, before deciding not to post it, UK MailOnline reported.

The reason why Facebook knows this information is because it can track what you type, even if you never post it.

The study, "Self-censorship on Facebook," found that men are more likely to "self-censor" their social network posts, compared to women, and this is especially the case if they have a lot of male friends. UK MailOnline reported.

Facebook data scientist Adam Kramer and student Sauvik Das studied the profiles of 3.7 million users by tracking the HTML form element of each page. This element is made up of HTML code that controls the boxes Facebook users type in to, including the status update box, according to UK MailOnline.

Scientists can track the changes in the code each time a user enters characters into one of the update boxes. Typing in the comment box on statuses, photos and other posts was also tracked by the researchers.

Facebook can't track the exact keys pressed, and it doesn't monitor keystrokes. This means the code doesn't reveal what is being typed. However, Facebook can track when characters and words are typed, how many are typed, and if the typed characters are deleted or abandoned.

The study found that 71 percent of the sample users typed out a post but never submitted it, and on average this worked out at 4.52 abandoned statuses and 3.2 abandoned comments, UK MailOnline reported.

Kramer and Das spent 17 days looking at 3.7 million profiles to track these "aborted status updates, posts on other people's timelines, and comments on others' posts."

Facebook users with a wide range and "diverse set of friends in fewer distinct communities" self-censor less than people with more "homogenous" friendship groups, the study showed. The researchers concluded that while 71 percent of users did last-minute self-censor at least once, they suspect, all users employ last-minute self-censorship on Facebook at some point.

"The remaining 29 percent of users in our sample likely didn't have a chance to self-censor over the short duration of the study," the researchers told UK MailOnline. "Surprisingly, however, we found that relative rates of self-censorship were quite high: 33 percent of all potential posts written by our sample users were censored, and 13 percent of all comments."

"The decision to self-censor also seems to be driven by two simple principles: People censor more when their audience is hard to define, and people censor more when they relevance or topicality of a 'space' is narrower."