Justice is blind--but can it be deaf?

People accept that it’s wrong to judge based on skin color, but most people think that there’s a right (read: white, middle class, “standard”) way to speak.

If I were reading this text aloud to you, your brain could use the sound of my voice—the way I combined vowels and consonants, the subtle shifts in inflection, the vocabulary I used, the order of those words, and maybe even the frequencies I produced—to determine my ethnicity. And your brain would very likely get it right.

Can you tell “what” I am from the slant of this typeface? I didn’t think so.

The field of sociolinguistics studies how language is socially constructed—that is, how who we are determines how we say what we say. And any good sociolinguist will tell you that, as listeners, we place almost as much value on the “how” as we do on the “what”: we judge people according to how they sound. Speech is as important as skin color to our assessments of—and potential discrimination against—people who are different from us. Sociolinguists call such discrimination linguistic profiling.



(Linguistic profiling can be used for more than just discrimination. Voices can be analyzed like fingerprints and used to ID individuals; such analysis can solve crime or open doors with voice recognition technology.)

But sociolinguists have a hard job. It’s nearly impossible to sort out all of the variables that contribute to our judgments about who someone is. Erik Thomas, Jeff Reaser, and Walt Wolfram at NCSU have spent parts of their careers trying to narrow it down. In a forthcoming chapter in Linguistic Profiling and Linguistic Human Rights, Thomas reviews several decades worth of scholarship, and the big conclusion is…a question mark. We know that people can make good guesses about ethnicity from phonological features—sounds alone, no syntax or vocabulary clues, but we don’t know which phonological features are the key.

Thomas writes that studies—most of which use recorded samples of African American (or “black”) speech as well as other control samples to survey listeners—have determined that there are prototypical features of African American speech. Sociolinguists call speech that contains such features “marked” speech. Studies also show that non-prototypical black speech (i.e. the speech of black speakers from isolated southern towns which shows more marked features typical of southern, not black, speech) is less recognizable to white listeners. But more prototypical black speech, even without marked words and syntax—is recognizable.

The key, says Thomas, is in the way that African American speakers use vowels and consonants and in the rhythm and intonation—or prosody—of their speech. But what’s the recipe? What features, when absent, erase the speaker’s race?

In 2004, Thomas and his colleague Jeff Reaser collected samples from African American speakers in Hyde County, North Carolina—a region where Black speech contains fewer prototypical features. They then used a computer program that allowed them to modulate vowel sounds to tweak, or monotonize, prototypical /o/ sounds. When they compared listener monotonized /o/s to the unchanged samples with “marked” /o/s, they found that the vowel helped listeners identify African American speakers. (A control group from the piedmont, where Black speech has many more prototypical features--sounds, words, and syntax--was much easier for listeners to identify.)

A follow-up study using different analytical methods and extending the listener sample to include West Virginia University students as well as NCSU students is forthcoming. Its results imply that listeners use different cues to determine the ethnicity of males and females.