February 29, 2004

Rolling and unrolling Indian r's

Thomas Friedman's column in today's NYT describes his experiences in an "accent neutralization" class in Bangalore that "teach[es] the would-be Indian call center operators to suppress their native Indian accents and speak with a Canadian one." He characterizes the phonetic content of the class with a passing reference to allophones of /t/ and /r/:

"Watching these incredibly enthusiastic young Indians preparing for their call center jobs — earnestly trying to soften their t's and roll their r's — is an uplifting experience, especially when you hear from their friends already working these jobs how they have transformed their lives."

"Softening their t's" is a plausible way to refer to the flapping and voicing of /t/ when it's not the onset of a stressed syllable. This is the process that makes "latter" and "ladder" homophones for most North Americans, and likewise can make "say 'fat' again" sound like "say 'fad' again".. This is an example of the class of processes that linguists call "lenition", which is basically a fancy word for "weakening" or "softening", though "softening" is not generally used as a term of art in the same sense. As a term of phonetic description, soft is sometimes used for voicing features (though sometimes voiceless sounds are called "soft" and sometimes voiced ones are), and sometimes for other features like palatalization (the "soft g" in George vs. the "hard g" in gorge). I haven't heard the phrase "soft t" used to describe the lenited allophones of /t/ in North American English -- but it wouldn't be a terrible choice of terminology, despite the danger of confusion with palatalization of /t/ in sequences like "what you" -> "whatcha".

But teaching young Indians to "roll their r's" in order to sound like Canadians? This is really puzzling, since there's a standard meaning for the term "rolled r", namely the kind of tongue-tip trill that most Spanish speakers have for word-initial /r/ ("la raza") or for medial /r/ written as a geminate ("perro"). I'm no kind of expert on the English dialects of our neighbor to the north, but I'd be willing to bet a substantial sum that few if any of them have any trilled r's. As far as I know, most Canadians have the same "bunched" r that most Americans do. This is a sound with a very interesting bit of acoustic physics behind it -- but that's another story.

Meanwhile, if the youth of Bangalore are really being taught to perform trilled r's as a way to sound Canadian, their future customers are in for a treat. More likely, though, Mr. Friedman (or his editor) is just confused. The local language in Karnataka state, where Bangalore is located, is Kannada, which has a trilled r. So perhaps the students in the class that Friedman observed were actually learning to un-roll their r's -- though I've visited Bangalore, and my impression is that its residents mostly used tapped r's in their English.

I'd be interested in a better-informed account of the mass-market Henry Higginses of the growing Indian call-center industry. I imagine that they spend more time on vowels and prosody than on consonants, but it'd be nice to know the facts.

[Update: Bill Poser suggests:

I have no information on what the call center training does, but I wonder if part of it doesn't have to do with removing retroflexion. The use of retroflexes in place of standard English apico-alveolars is a salient characteristic of Indian English as well as of the loan phonology of the Indian languages that I know about. The retroflexes are referred to as "hard" in the non-specialist literature, e.g. in discussions of character sets by computer people. So "softening" might be removal of retroflexion.

This is certainly plausible. Or maybe Friedman just registered that the students were learning how to adjust the pronunciation of various sounds; wanted to be more specific because it makes for a more engaging and readable account; picked "t" and "r" at random because they're common letters whose English pronuncations is somewhat regular; and threw in "soften" and "roll" by process of association.

All that I can confidently conclude from what Friedman wrote is that he's not especially interested in phonetic description, which is hardly news. As for what and how Bangalore call-center recruits are really taught about how to imitate various English dialects, those are interesting questions that will have to answered by someone who knows the facts.]

Posted by Mark Liberman at February 29, 2004 09:56 AM