The Secret Life of Pronouns (11 page)

The bottle project influenced the way I think about personality. As more and more people described the bottle, Cindy and I started to analyze their descriptions in new ways. Initially, of course, we linked people’s function words with what we knew of their personality. For example, the more people used pronouns (as in “I hope you’re not thirsty”), the more sociable they reported being. I-words were used slightly more by people who were insecure, anxious, or depressed. People displaying markers of formal thinking used high rates of articles and prepositions in their descriptions and tended to be older, more organized, and conscientious.

But something was missing in these analyses. Go back and look at the examples. Yes, the three people use function words differently but what is more striking is
what
each pays attention to. Why does one person notice the colors and another imagines what the bottle would feel like in his hand? How people naturally describe an object tells us a little about the ways they think and perceive. For example, when I did this task on my own, it never occurred to me to describe how the bottle would feel.

We needed a more efficient way to analyze how people were seeing and experiencing the picture. A few weeks later, Cindy stopped me in the hall with a solution. It was simple and ingenious. The technique, which eventually became known as the Meaning Extraction Method, relied on the most common content words in all the bottle essays we received.

To understand how the technique works, look again at the three bottle texts. As you would predict, the most common words are function words:
the
,
it
,
is
,
a
,
of
. In addition, there is a much smaller number of content words. Although the three texts consist of a total 110 different words, within the 40 most frequently used words, there are 14 content words:
water
,
bottle
,
red
,
small
,
white
,
cap
,
clear
,
easy
,
empty
,
label
,
left
,
letters
,
light
, and
plastic
. If there were, say, almost 1,500 complete essays that people had written about the bottle, there would be far more content words. In fact, we found that 1,500 essays produced over 200,000 total words, but only about 175 content words were used frequently.

Now imagine that we go into each of the 1,500 essays and determine how often each person uses each of the 175 common content words. Next, we determine how these content words naturally clump together. If you think this sounds suspiciously like factor analysis, you would be correct. Factor analysis, as you recall, can mathematically determine which groups of content words are used together by the various authors of the bottle texts. We discovered that a person who uses the word
yellow
in an essay is also a person who tends to use other color words—
blue
,
green
,
white
—as well as words like
sky
,
mountain
,
banner
, and
label
. When you look at this clump of words, you realize immediately that these are all words people typically use together when describing the label on the bottle.

Another word clump includes words like
cylinder
,
shape
,
cone
,
top
,
tall
,
wide
, and other words related to the bottle’s shape. Yet another theme that pops out includes words like
light
,
gray
,
shadow
,
reflection
,
background
,
wall
, and
table
: a background/lighting theme. There is even a theme that identifies people who are obsessed with whether the cap’s seal has been broken and how it would feel to hold or drink from the bottle—much like the third essay above.

If you are not an expert in computers and factor analysis, you are probably thinking, “Duh! Why do you need a computer program to tell if someone is talking about colors versus shapes? It’s obvious.” Trust me, it isn’t obvious to a computer. The Meaning Extraction Method is almost magical—it can automatically pull out the themes that exist in a large number of similar text files. The computer—rather than human beings—can detect the underlying themes in people’s writings.

From a psychological perspective, it allows us to see what kind of people emphasize which themes in their writing. Even at the most superficial level, the results tell us that there are some people who focus on colors and others who don’t. Another group writes about the words and lettering—and others don’t.

We have also explored the behaviors of people who emphasize different themes of the bottle. The most interesting are those who write about the lighting and shadow cast on the left side of the bottle. Among college students, those who write about the shadow tend to be more thoughtful and artistic and less concerned about appearances. They make higher grades, go to more art shows, and play more computer games. They also are less likely to drink heavily, to make to-do lists, to vacuum, to fix things when they break, to take baths (although, fortunately, they shower as much as others), and to blow-dry their hair.

Other patterns emerge as well. Those who emphasize the words on the label are more likely to be female and to read more. Those interested in the texture and the feeling of the bottle tend to report more physical symptoms of all types. And the students who write about the surface of the bottle seem to erect a barrier between themselves and others. They report being less agreeable, avoid having heart-to-heart talks with others, and, if they have a major upheaval, prefer not to talk with others about it.

Who would have thought that describing something as inconsequential as a bottle of water could reveal so much? Although promising, the patterns of effects are not very strong—although they are statistically trustworthy. In other words, it would be foolhardy to ask our friends to describe the bottle and then rush out and buy some of them a blow dryer if they mention the shadow in the picture. Instead, think of these findings as a hint of what is available to us in analyzing language. Imagine what we might see if we ask people to describe more complex things.

SEEING YOURSELF BY SEEING OTHERS

I love the bottle test. Buoyed by the findings, my students and I tried another picture. This time, we wanted a more complex and socially relevant stimulus. A regular picture of real people was needed. To find one, I rummaged through my desk drawers looking for pictures. Presto—I discovered a photo of two very dear friends at an outdoor party my wife and I had had several years earlier. With their permission, I started using this picture in several studies.

Before going into the findings of the Garden Party Project, look at the picture above. Spend a few minutes writing down how you would describe it to another person who couldn’t actually see it. If you decide not to write your description, at least do the project in your mind and remember what you focus on.

Perhaps more than anyone, this picture fascinated one of my graduate students, Kate Niederhoffer. Kate’s research interests spanned traditional social psychology and the world of business and marketing. She looked at the picture and wondered if the ways people described it could predict their buying patterns. More broadly, she wanted to know if people’s daily behaviors were reflected in the ways they looked at the world. Kate, who now helps run a marketing company that uses text analysis, set up a website where hundreds of people from all walks of life described the picture and also completed an exhaustive questionnaire about their daily behaviors. (A short version of the questionnaire, the LIFE scale, can be completed at www.SecretLifeOfPronouns/LIFE.)

Hundreds of people ended up describing the backyard party picture. The types of descriptions varied tremendously. Some examples:

This is the woman’s house. She is happily married. Right now this guy is telling her a story or a dirty joke because the woman is obviously surprised by what he is saying. She is drinking home made iced tea. These two people are not married to each other but both of them are married.

—22-year-old female, recently engaged to be married

The lady just came from inside the house after pouring herself some beer from a freshly tapped keg. She has already had a lot to drink making her actions seem artificial. The man looks like he is more interested in grabbing the beer from her hand rather then meeting her. The man wants to grab the beer from her hand and take off.

—19-year-old male who reported drinking alcohol excessively

There are two people talking in the backyard of a house where a barbecue party is being held. She is wearing a long sleeve, denim, button-up shirt (probably from the Gap), tucked in, with the sleeves rolled up twice onto her forearm. She has medium brown hair that is tastefully cut (with layers) to just under her chin. She also has on a black (Gucci?) belt and a red bracelet on her right wrist.

—25-year-old female who spends over $500 per month on her wardrobe

The two people in this picture are engaged in a deep discussion about politics. The woman is expressing her surprise that the man is a Republican. She is shocked at the logic of his argument … The man is demonstrating (with his hands) that he appreciates the ideal of “smaller government” that he feels is best attributed to the Republican party.

—21-year-old female who is active in politics

The Meaning Extraction Method pulled out a number of consistent themes from people’s writing. Common ones included the social connection between the two people (are they dating, married; what has he just said), an analysis of their clothes (from simple descriptions to fashion observations), accessories (their watches, his glasses, her lipstick), what the woman is drinking (alcohol, tea, is she drunk?), the yard (flowers, trees, the trampoline), the house (window, siding, roof), and more about the two of them (her facial expression, their races and ages). As with the bottle experiment, each of these themes was actually a clump of related words.

As you can see, the themes people wrote about were indeed related to their own lives. People in dating relationships, especially women, wrote about the couple’s likely relationship and often commented on the wedding rings they apparently are wearing. Participants who reported spending a great deal of money on their own clothes and clothing accessories were much more likely to write about the couple’s clothes and accessories. Gardeners commented on the flowers and trees. And the more people admitted to drinking alcohol excessively, the more likely they were to infer the woman was drinking beer and was a bit tipsy.

That word use can predict buying behavior is not news to Google and other search engine companies. If you use a product like Gmail and receive an e-mail from a friend about his having Raisin Bran for breakfast, you might notice that a number of discreet ads appear on your e-mail page about fruit snacks, Mango Fruit Chillers, or a drug to reduce cholesterol. Today I received an e-mail from a family member whose forty-year-old friend has been irrational and moody and was asking my advice on what to do. The ads that popped up included “Treat Teen Acne,” “The Best Book for Teens,” and “Have a New Baby?” Apparently, some of the words—perhaps words like
irrational
and
moody
—are statistically associated with teenagers and having a new baby.

PROJECTIVE TESTS

Using pictures of the bottle and the garden party is a variation on a much older idea in psychology. Sigmund Freud, Carl Jung, and especially Freud’s daughter Anna Freud claimed that people naturally project their own thoughts and feelings onto other people and objects. Someone who may be angry at themselves for missing an appointment may run across an old friend and comment, “What’s going on? You seem to be a little angry about something.” Through the defense mechanism of projection, the individuals may not recognize anger in themselves but see it everywhere else.

In the early 1920s, a young German psychoanalyst, Hermann Rorschach, expanded on this idea. He developed a method in which people looked at ambiguous pictures of inkblots and then described what they saw. The Rorschach test was based on the idea that people’s deepest emotions and concerns would be projected onto the inkblots. For example, a person who lived in constant fear of an abusive parent might see one of the inkblots as a ferocious bear, whereas another person in a happier family might see the same inkblot as a butterfly. Since Rorschach’s early work, thousands of therapists have relied on the Rorschach test to explore people’s psychological states. Many researchers question the Rorschach because people’s responses are not reliable and vary a great deal depending on the therapist who administers the test. I can’t help wondering if the method might be improved by administering and scoring the Rorschach by a computer.