What Do Men And Women Message About?

What The Link Between “Constipation” And “Shopping” Can Mean For The Future of Social Science Research 

Men and women tend to talk about different things. Here are the basic differences, according to some of the latest research:

If we believe language is a window into people’s thoughts, then studying word usage can lend insight to how different groups — for example, men and women — think differently.  Fortunately, with the growth of digital media, researchers have been able to collect language samples on a much larger scale than ever before. 

In 2013, in the largest study of its kind, researchers at the University of Pennsylvania and Cambridge University examined differences in word usage between females and males from 75,000 social media accounts.  Below are the word clouds from the study itself.  The word groups on each side are ones that are used disproportionately more by that gender than the other.  The larger the word, the more it is used by different people of that gender and the redder/more colorful it is, the more frequent the word is used. We took the liberty to group each into the broad categories at the beginning of our article.  

FEMALE

MALE

We at Mei decided to conduct a similar study on text messages from about 45,000 English-language texters, 59% males and 41% females. While our sample size is smaller than the UPenn study, the total amount of words is twice the size (approx. 600 million words vs. 309 million in the UPenn study).

Our study bubbled up “feminine” and “masculine” word categories that were largely in line with UPenn’s categories, despite their study looking at social media posts while ours focused on text messages.  We found that women use cheerful, emotion, domestic-work/home, and anticipation words approximately 15% more than men do. Similarly, we noted that males used words related to the military (1.54x), technology( 1.38x), politics (1.29x), and money (1.13x) more. However, there was one notable difference. The UPenn study did not reveal significance difference between females and males in one major category: Health.  We found it was one of the most distinctly “feminine” categories of words.  

Below is a table of the top ten term categories used more by women vs. men and vice versa.

We found women used health related words 30% more, surpassed in “femininity” only by fashion-/fabric-related words.  Health-related words are even more prominent than beauty words (like makeup, lipstick, perfume, and jewelry), usually thought to be feminine. 

Conversely, in the UPenn study, health words were in the bottom half of the categories they used.  The omission of such a stark difference between the two sexes suggests health topics tend to be excluded from social media, where people often portray a more “public” version of themselves.  By analyzing text messages, we’d argue our study portrays a more authentic version of people.

Methodology

The UPenn study classified words into 64 pre-defined word groups, defined by the Linguistic Inquiry and Word Count, or LIWC. Our study classified words based on Stanford’s Empath tool, which has been shown in research to be highly correlated with the LIWC. Empath classifies words from our dataset into 194 categories.

The tables above show term frequency (the number of times a word was used divided by the total number of words used) whereas UPenn’s study used document frequency (the number of people who used a word divided by total number of people whose words are captured).  We did not discover any significant difference in our results between measuring term frequency or document frequency.  Because term frequency results can be skewed by a small number of people using a term very frequently, we display document frequency results in the remainder of this article. 

Out of the 157 health-related words as classified by Empath, approx. 85% (132) were used more commonly by women than men. Below is a table of document frequency differences for the top ten most used (i.e. term frequency) health words sorted by total word count.

Who would have guessed the words ‘diarrhea’ and ‘constipation’ were texted disproportionately more by women than by men?  These words are about as feminine as ‘shopping’ and ‘hair’!  The table below highlights the words we found to have the largest skew towards women.  Many of these words aren’t in the Empath word list and we’ve included words from other categories for reference. 

We don’t immediately know why women text about health much more than men.  Perhaps women are more comfortable discussing topics of health or health tends to come up more when discussing beauty and relationships (which we see above as decidedly more “feminine”).  Perhaps it’s because women are more often the caretakers of others.  Or perhaps women just experience more health issues than men.  (If anyone has insights on this please, please feel free to share).  Whatever the cause may be, the finding highlights that people behave very differently on social media than they do in private text messages.  

Our Public vs. Private Selves:  Social Media vs. Text Messages

Social media posts tend to miss what people don’t want to share publicly and are subject to what’s called desirability bias.

In social science research, social desirability bias is a type of response bias where survey respondents have a tendency to answer questions in a manner that will be viewed favorably by others.  It can take the form of over-reporting “good behavior” or under-reporting “bad,” or undesirable behavior.  This bias has been a perennial problem in social science research.  Any experiment that has relied on surveys or examining people’s written words (blog posts, poems, books, articles, etc.) has been subjected to desirability bias.

Nowhere is desirability bias more prominent than in social media, where people tend to exaggerate the positives in their lives and under-report the negatives.  Messages like “running late for my dentist appt” and “goddamn this diarrhea” aren’t newsfeed-friendly the way “thanks for all the birthday wishes” and “loving this adorable puppy” are.

Messaging is by far not immune to this bias, but it offers a level of authenticity and insight to what people do or say that can elude mediums like social media posts and even internet searches. Text messages offer a peek into how people communicate privately to those closest around them. They are also far more voluminous than social media posts (average person sends about 50 texts a day vs. say 2 social media posts).

Implications for Social Science Research

Studying messages, a core mission of ours, is in its infancy in terms of contributing to social science research.  We wrote this article to illustrate how the sharing, studying and publishing of social data can be a starting point for unlocking knowledge.  So long as data is used responsibly and the privacy concerns of the people sharing it are respected, we truly believe messaging data can quickly uncover mountains of knowledge about people in general — while benefiting users, researchers, and platforms like ours designed to better understanding communication.

We’re working on this and are also looking for partners to work with us, responsibly.  In the meantime, we will be publishing future articles diving into topics we raised here and elsewhere.  Stay tuned.