When consumers whisper their values

Editor’s note: Peyton Mason is president of Linguistic Insights, Inc., a Charlotte, N.C., research firm.

When people interact online, all they have is words. They cannot see facial expressions or hear shifts in tone or emphasis. They don’t spend much time typing, and they don’t use many words but they make those words work. They project their opinions: We should do this…You wouldn’t do that. They hedge their bets: Well, maybe I could…Sometimes they might. They specify conditions: If this happens, then…

Browse a transcript from an online focus group or an online chat, and you will very likely be struck by the leanness of the communication: short one- and two-syllable words, lack of complete sentences, missing subjects and verbs. This terseness leads people to think that online dialogue conveys far less information than face-to-face conversations, and it does. But people’s word usage in the online environment goes a long way to make up for the loss of cues from body language and intonation.

  • The online universe does not just look or feel different, it is different. People change their language use, or modify their ways of cooperating with each other, and sometimes even change their names. But people’s inner motivations don’t really change. And they keep the same style they use when they talk. That style can be captured.

Even in the online universe, as in face-to-face interviews or focus groups, style can be scaled to measure group solidarity or cohesion; style can also be scaled to measure consumer perceptions of a brand or service. People have a constellation of preferred ways for expressing themselves, each dependent upon the context of the conversation, and they take those preferences online.

An analysis of consumers’ online styles can show the social engagement that is occurring within groups or between an interviewer and interviewee. Style, even online style, can be scaled to determine the strength of a stance or position that an individual is taking toward a topic.

Our research finds the meaning that’s wrapped up in people’s style. Ask people’s opinion about something. Here’s what they typically use their own style to do:

  • tell you what they don’t like;
  • muffle their response, when they want to avoid hurting your feelings;
  • disclose personal experience when they’re seriously interested;
  • qualify their reaction when they’re feeling cautious.

It’s the way they do it — their signature style — that tells you what they mean. People’s style is located in the way they choose and arrange their words. Our approach looks at those words from multiple perspectives, combining quantitative and qualitative approaches to consumer communication to uncover their points of view, including the ones they thought were hidden.

Multidimensional approach

To get under the surface, we first use a set of multidimensional techniques that are typically used to analyze a collection of spoken, written or online texts. Sometimes the analyst uses these techniques to look for potential areas of deception. Other times, an analyst uses them to look for evidence of authorship. We use these techniques to look for ways people signal their attitudes.

We draw on statistical analyses to determine underlying associations across a speaker’s — or a group’s — language features. The approach is based on the assumption that “statistical patterns reflect underlying shared communicative functions” (Conrad and Biber 2001:7).

The approach is multidimensional because style is considered a bundle of language features that co-occur in text because they work together to mark some common underlying language function.

The qualitative aspect of this analysis is the interpretation of how the factors signal the impact of a particular bundle of language features.

What we quantify

We employ a database of conversations. Each new transcript from an online focus group is compared to our database of online styles. We quantify and analyze consumer language, in order to determine:

— how comfortable the participants are in the communication situation;

— how tentative they are with their personal response;

— when and how they qualify their opinions;

— when they’re ready to commit to something;

— when they’re edging away;

— how they recommend action; and how, all the way through, they send soft signals of their own values.

Our database of language styles contains 750,000 words and includes both face-to-face and online conversations (chats), interviews and focus groups. It’s important to have selections from numerous topical areas. For instance, in the online universe of chat groups, topics serve as “places” where language styles can differ in the same way the different sections of a high school — classroom, auditorium, lunchroom, gym — constrain the way teens choose their words, their tone of voice, even the choice of who speaks to whom. To ensure a broad base of language usage, the full database contains groups and chats on: travel, family, friendship, money, music, religion, friendship, health, hobbies, books, sports and politics. The subset of online focus groups covers finance, travel, fashion and online retail, representing different regions, ages and genders.

From words to dimensions

We’ve all heard the old saying: “It’s not what they say; it’s how they say it.” To see how they say it, we coded all the texts in the database. Each text in our full database was coded for 90 parts of speech, which were then analyzed for frequency. In this part of the analysis, we look at how people use their words to take a stand on a topic.

Our series of factor analyses identified the 33 strongest patterns of word choice. Next, we scaled the factors to establish benchmarks that locate when members of a focus group change their position, signal their commitment or move into evaluation. People’s stance shifts as the words scroll down the screen or arrive on the bulletin board. Meaning is not found in words alone. It’s also in how words are put together. We measure both the what and the how. Here is why: In online focus groups, responses are short — five to seven words is a typical utterance. Two lines (or about 18 to 20 words) are a lot to type. Responses are time-constrained and pressurized; utterances are unplanned, compressed and full of typos. And the responses are truncated: The participants do not repeat nouns and verbs used by the online moderator.

Our benchmarks signal what people have to say, when they say it, and how people collaborate to construct meaning when they are online, regardless of whether they are chatting, sending e-mail or participating in focus groups. The “when” benchmarks are the locations within a focus group discussion in which the participants are taking a stand. The “how” tells us what the stand is that they are taking.

The participant in online discourse must be able to realign in seconds to changes in topic, addressee-addressor relations, size of audience in terms of number of participants and the entrances or exits of conversationalists, and in the tone of general or specific interactions. What the moderator or analyst needs is some way to go beyond the obvious, identifying subtle shifts and masked or buried “hot areas” of a transcript that typically stay hidden until the fifth or sixth pass through the material.

In online focus groups, people don’t hand you their meaning with the verbs they choose. Half the time, they leave out the verbs. Instead, meaning is carried by: adjectives, adverbs and adverbially-used prepositions — not all at once, and not to the same degree. But they are the carriers of meaning, and participants arrange other types of words around them — sort of like trimming a tree — to signal the direction of the online discussion and the strength of their feelings or opinions. When scaled, the interactions among these carriers can identify the successive stages of the stance a person is taking toward the topic under discussion.

For example, in the online universe, personalization is handled by first-person pronouns, but the second-person pronouns are typically used as a way to project belief or action, one step away from “I.” Combinations of verbs of perception or cognition with modal auxiliary verbs — we called them “helping verbs” in school — can signal conditional interest.

Stance Analysis Monitor

The Stance Analysis Monitor methodology lets us identify consumers’ feelings toward a topic or brand or other stimulus, and their indirectly-expressed opinions. Focus group participants do not always want to say directly that they value or more often, what they do not like about a product or brand. Sometimes, they’re not sure of what they think until the words spill out of their mouth or onto the keyboard.

Stance is measured along four dimensions that measure the degree:

  • Strength — this is the intensity with which people whisper, speak or shout their position.
  • Conditionality — this is the extent to which people will qualify or list conditions constraining an action or a response.
  • Hedging or waffling — here’s where people back off, mitigate or soften their position.
  • Action-orientation — here’s where people announce their intended action or project that action onto other people.

No single grammatical or semantic feature tells the story. It’s the interaction between content and context that determines which language features are used. Stance Analysis Monitor is multidimensional, and each dimension signals informational load and social involvement (Conrad and Biber, 2001).

The impact of language combinations

Language choice can signal social involvement. Ever read a note — or an online chat — and feel like you can spot who knows whom, and how well? Some grammatical forms suggest a speaker is focused on becoming involved with the listener: When speakers delete the “that” after a statement of belief, opinion or perception, they are often suggesting or initiating a social relationship, increasing the potential for social cohesion, and inviting a similar response. “I believe we’re ready to get that” has a different impact from “I believe that we are ready to get that.” And “You’ll think it’s great!” is far more comfortable with its exclamation point than “You’ll think that it’s great!” Using “that,” particularly online, gives a deliberate, weighty, formal air to the conversation. If your online focus group starts using “that” in its sentences, watch out: participants may be feeling uneasy, or may be backing away from commitment.

What about “you”? “You” doesn’t always mean “you, Dear Moderator,” or “you, Dear Reader.” It is often used as a sort of generic, where it suggests informal, colloquial, familiar conversation among people who are acting as if they ought to know each other. The “you” in these examples from our collection is not the “you” of direct address to the reader or hearer: here, the “you” means “I, we, or everybody”

You definitely need comfort in shoes.

You can get them like at Goody’s for only $5.00.

Dressing up makes you feel good.

The trick is to decide when you means I, we, somebody, or anybody. The answer is usually keyed to context. By context, we mean who’s talking to whom, about what.

“You” is often used to suggest a hypothetical situation. Its use lets the speaker/writer achieve a little distance, moving away from full ownership of a situation or idea. That lets the speaker/writer make a comment on a situation without taking full responsibility for it. Communications researcher H-Y Tao thinks that the first person to use “you” in this way in a conversation is actually taking on the status of the privileged speaker. We think it’s a gentle form of one-upmanship. Look at how the writer assumes authority by using “you” in this sentence:

“If you’re going to write ads aimed at your audience, you’d better use their current language.”

What does it all mean?

A multidimensional analysis of online language spotlights the word-use equivalent to the visceral responses, body language and involvement that are not observable online. It bridges the gap of being able to “view” participants as you would for an in-person group. You do not have to hire a brass band to take a stand: most people do it with a wink or a single word. Finding the online equivalent is where an approach like Stance Analysis Monitor is useful. It helps to get under the surface to detect what is important. It:

  • Replaces the loss of body language and intonation of in-person groups.
  • Spotlights participants’: opinions and rationales; personal engagement with the topic and each other; waffling; and potential for action.

After the focus group or the interview is over, the analyst has three tasks. Everything rests on how he or she takes an inventory of the changing positions, changing degrees of commitment, changing attitudes throughout the group or interview. The first task is to figure out the big story of the ebbs and flows in a discussion. The second is to look objectively at what people meant, when alliances or topics shifted, when the meanings blurred, and how engaged the participants were. The third is being able to compare transcripts, and that’s hard.

References

Conrad, Susan and Douglas Biber, eds. 2001. Variation in English: multi-dimensional studies. Harlow, London & New York: Pearson.

Tao, H-Y. 1998. An Interactional Account of the Impersonal Pronoun You in Conversational English. Paper presented at IPrA conference. Reims, France.