May 05, 2006


Passed on to me by Joe Clark, a piece from the 4/22/06 Toronto Star on Mitsubishi Canada, available here under the header "Lancer, Eclipse lead Mitsu return in Canada", in which it's claimed that many Canadians pronounce the brand name "Mit-soo-BOO-shee".  To judge from the enormous number of "Mitsubushi" spellings I googled up on 4/24, it's not just Canadians; the spelling's used by English speakers all over the world.

Eventually, I checked out all sixteen ways of distributing the vowel letters I (representing [i] or [I]) and U (representing [u] or [U] or sometimes maybe a reduced vowel) across the four vowel positions in the frame M_TS_B_SH_.  Of the 15 possible misspellings, all except two (U I U U and U U U U) are attested, but they come in four clusters.  I U U I (Mitsubushi) is on top, along with two close contenders, then two more at about half or a third the frequency of the big guys, then a cluster of low-frequency spellings and one of very-low-frequency spellings (and then, of course, the two total losers).  Even more eventually, I found a way to make some sense of these differences in frequency.

First, the Star piece, by Jim Kenzie, which begins:

   NEW YORK CITY - In a world where brand name seems to be so critical to success, Mitsubishi is fighting a multi-layered battle.

   First, in a word-association game, few Canadians recognize the name at all. Second, many who do, pronounce it Mit-soo-BOO-shee, for completely inexplicable reasons.

   Nobody calls it Toy-OO-ta, do they?

   Third, if consumers do know the name Mitsubishi, it is more likely to be for a TV or VCR than a car except for video game freaks, for whom the Mitsubishi rally car, the Lancer Evolution, has its own mythical status (and is, as yet, unavailable in Canada).

As I then reported to the ADS-L, googling on Mitsubushi pulled up a startlingly large number of raw webhits (on the order of a hundred thousand) -- from the U.S., Australia, the U.K., and India, just within the first ten hits.  Then eventually a pile from Canada, South Africa, New Zealand, and Ireland.  As far as I can tell, mostly with reference to cars, but sometimes with reference to electronic equipment.  This is a stunningly common and widespread misspelling.

I then moved on to the other logically possible misspellings, first just using raw webhits, then going back and using the number of hits at the point where Google tells you that the rest of the hits are similar to the ones already displayed.  The two counts gave essentially the same clusters and rankings, with only small differences between them.  From here on, I'll use the reduced figures.  And the results are...

Top dogs:
    1.  I U U I (796), with U spread to position 3
    2.  I I U I (767), with U moved to position 3, in exchange for I
    3.  I I I I (614), with all Is

Moderate frequency:
    4.  U U I I (322), with U in position 1
    5.  I U I U (178), with U in position 4

Low frequency:
    6.  U I I I (61)
    7.  I I I U (41)
    8.  U U U I (31)
    9.  U I U I (29)
    10.  I U U U (26)

Very low frequency:
    11.  U U I U (6)
    12.  I I U U (4)
    13.  U I I U (3)

Total losers:
    14.  U I U U (0)
    15.  U U U U (0)

(One blogger reported, a propos of #9: "Then came Mitsubishi, which for some reason Israelis call Mutsibushi. and in 1990 came Honda, but only American made Hondas (from Ohio), and then the rest."  I haven't pursued the claim that Israelis are given to variant #9.)

[Addendum: hint and you shall receive.  In no time at all, Aviad Eliam e-mailed me with news on the Israeli-Japanese front: "Unlike the blogger, however, Mutsibushi didn't sound to me like the common error, and my Israeli friends concurred. They suggested IUUI and IIUI, two of the top three contenders in English. I went and did my own quicky google search in Hebrew" -- and found (in raw webhits) English #1 I U U I on top (2130), then #2 I I U I (948), then a drop to #8 U U U I (189) and #4 U U I I (152), another drop to #3 I I I I (37), #6 U I I I (28), and #9 U I U I, i.e. Mutsibushi (17), one each for #10 I U U U and #5 I U I U, and no hits for the rest.  Roughly comparable, especially at the top, with the English data.  And, though Israelis occasionally do write (in Hebrew, of course) "Mutsibushi", they don't do it at all often.]

At this point, I invite you, the reader, to pause and speculate about what might be going on here.  Why these clusters?

While you're thinking, I'll argue that the problem isn't nativization of foreign words.  To begin with, hardly any of the writers on the web will actually have heard "Mitsubishi" as pronounced by a speaker of Japanese.  What English speakers have to go on is the spelling of the word, the knowledge that it's a Japanese name, and productions of the word by other English speakers.  Even just given the first two, the word is perfectly easy to nativize into English, with the accent pattern 2 0 1 0 (where 1 is primary accent, 2 is a weaker accent, and 0 is unaccented), with the T and the S of the English spelling split between the two first syllables, and with "furrin spelling" values for the vowels (a high front (unrounded) vowel for I, a high back (rounded) vowel, or possibly a reduced vowel in position 2, for U).

Instead, my hypothesis is that the word presents a difficulty for memory and recall.  It has all Is, except for one syllable, and that syllable is the least salient one in the word (accented syllables are the most prominent -- here, the third syllable has the primary accent, and the first syllable a weaker accent -- and the first and last syllables of a word are, in general, also salient, but the second syllable has no kind of salience).  So: how to reconcile the specialness of the U in the second syllable with the lack of salience of this syllable?

One way involves remembering that the second syllable has U in it, but spreading that vowel onto the immediately following, accented, syllable, where it can stand out: I U U I (with the Us in the middle and the Is at the edges, in a nice pattern) instead of I U I I.  This is #1, Mitsubushi.

Another way is to move the U into the accented syllable: I I U I, #2.  This is a bit less faithful to the original I U I I than #1 is -- it differs from the original in two positions rather than only one -- but from a psychological point of view, it's a very likely error, since it involves recalling (correctly) that there's only one I, which is, however, misplaced by appearing in the most prominent position.

Still another way is just to forget about the U, and use all Is: I I I I, #3.

Put another way: with #1 you remember that U is important to the spelling, but get more than one; with #2 you remember that there's exactly one U, but get it in the wrong place; and with #3 you don't remember the U, because it was off in a corner in the first place.

These three solutions to the U problem are somewhat parallel to the misspellings that result for words that have one doubled and one single consonant letter in adjacent (as in corollary) or parallel (as in  [Amy] Gutmann) positions.  One solution is remember that there's doubling and go all the way with it: corrollary, Guttmann.  Another solution is to remember that there's exactly one doubled consonant letter, but misplace it: corrolary, Guttman.  (For the name of the president of the University of Pennsylvania, I'm especially prone to this error; unfortunately, I've had occasion to cite her work, and I don't always get it right.)  Still another solution is to forget the doubling: corolary, Gutman.

So the top dogs make a lot of sense. 

The moderate-frequency errors preserve U I in the middle of the word, but use a U in one or the other of the two secondarily salient syllables: the first syllable, in U U I I (which has the Us first, then the Is), #4; or the last syllable, in I U I U (which has a pleasing alternating pattern), #5.  It even makes sense that #4 should lead #5, since the first syllable is doubly salient (by virtue of position and also accent), while the final syllable is salient only by position.

The third cluster mostly has patterns with only two of the four model vowel slots preserved, plus one pattern, #9 (U I U I), with only one preserved (though it has a pleasing alternating pattern).  The other patterns at the bottom of this cluster -- #8 (U U U I) and #10 (I U U U) -- have three Us in them, which takes them very far from the model.

The fourth cluster has one pattern with two of the four model vowel slots preserved, but three Us, and two patterns with only one model vowel slot preserved.

Finally, you get the pattern that preserves NONE of the original vowels (U I U U), and the pattern that has no Is at all (U U U U).  These are desperately far from the model, and, fittingly, they get no (legitimate) webhits.

Overall, we see relatively frequent errors that make sense in terms of what people are likely to remember about words and use in recall, followed by much less frequent errors that mostly look like random stabs, but nevertheless are ranked roughly in the order of their distance from the model, in terms of number of vowel slots preserved and number of Is.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at May 5, 2006 04:18 PM