April 23, 2005

Unreliable dialect identification help

The dialect test available if you click here is based purely on words, and in some cases basically slang items (question 10: what do you call an easy course, a crip course, a gut, or a blow-off?). You can decide whether it does a good job of classifying you, but my results were:

Your Linguistic Profile:

40% General American English
35% Yankee
15% Dixie
5% Midwestern
5% Upper Midwestern

Well, the truth is that I'm originally a speaker of middle-class Southern British, with some phonological and lexical features modified over the past 25 years by living on the West Coast of the USA. I've spent no time in Dixie; hardly any in New England; a little in the northern Midwest. The excess of alleged Southern features over Midwestern is hogwash (oh, all right, maybe that's a word from a dialect other than my original one; let's call it balderdash).

Now, you might say that my dialect is confused by my emigration, and it misled the test. But my friend Gerald Gazdar is not confused. He speaks a very cultured 100% Southern British English with no American features whatsoever (he couldn't supply any answer for question 10 at all: they don't have easy courses in Britain), and he got a report of 45% General American English, 25% Yankee, 20% Dixie, 5% Upper Midwestern, 0% Midwestern. Gerald Gazdar one fifth pure Dixie? Y'all bin drinkin' moonshine or sump'n?

So the test is highly erratic when used on British English speakers. Perhaps that's an unfair criterion. Those born and raised in America who have definite ideas about their dialect can check things out for themselves, but in general I'd say it will be purely for personal amusement; it's not science. Look at the territory covered by the test's 20 questions:

Your Test's Profile:

75% presence of selected geographically variable lexemes
25% phonology/phonetics of particular lexical items
 0% general phonological rules
 0% morphology
 0% syntax

I doubt that's a large enough or diverse enough portfolio of test items for a standardized test of local dialect in a community of over 280,000,000 speakers who have been spreading over a continent for four hundred years.

[Added later: It has been pointed out to me by Wes Meltzer that the test appears to be based on the dialect survey devised by linguist Bert Vaux of the University of Wisconsin, Milwaukee; see this page. And in defense of the test, John Cowan writes to say:

Well, of course a test of American English dialects is going to cough up a hairball when taken by people whose English comes from a different dialect group, even if they have some overlay from living in this country a while.

I'm from New Jersey, just outside the New York City isogloss bundle (I'm rhotic, e.g.), and my parents (who can influence your lexical choices if you are fairly isolated from your peer speech community as a kid, what in AAVE and Labov is called a "lame") were from Detroit and Philly. Bingo:

55% General American English
35% Yankee
10% Upper Midwestern
 0% Dixie
 0% Midwestern

Couldn't be more on-target.

Point taken. Try it yourself. But keep in mind the sharp limitations: it's assuming you live in the continental US and have the features associated with your region as assessed by fifteen lexical items and five phonological properties.]

Posted by Geoffrey K. Pullum at April 23, 2005 05:28 PM