Madagasc + ?

« previous post | next post »

Yesterday, in explaining why he didn't open his post "Half golliwogs and other UK linguistic news" to comments, Geoff Pullum wrote:

"I'd rather eat a live Madagascan hissing cockroach than see a hundred comments on the above."

My reaction was to wonder "Madagascan?" I always thought it was "Madagascar hissing cockroach", with the simple place name used as a modifier. A quick check verified that Geoff's version is indeed in the minority, 3,180 to 24,200, though not nearly by a large enough factor to explain my confusion. (It also turned up the image reproduced on the right, which suggests that there are some people out there who might actually enjoy eating Madagascar hissing cockroaches.)

Anyhow, I realized that I wasn't quite sure what the derived adjectival form for Madagascar really is, except for Malagasy, which I thought of as being specialized for the language. (Though it isn't, apparently.) The OED has Madagascan, Madagascarene, Madagascarian, and Madagascrian, with all of them linked to Madagascan, which links to Malagasy as well, and also Malagache. On the web, Malagasy beats Madagascan, 5.3 million to 266 thousand (though it would be a big mistake to conclude that "Malagasy hissing cockroach" is the favored form for the insect pictured above).

I believe that the other alternatives are all less common, though Google claims to find 21 million hits for Madagascarene. I'm skeptical of this figure — Yahoo only finds 351 (that's three hundred and fifty one, with no zeros after it), and MSN finds 37. Google's extravagant count in this case seems to be the result of over-enthusiastic spelling correction.

Madagascarese is out there on the web as well, but not in impressive numbers or contexts.

Why the variants? As the OED explains, it's partly thanks to the Madagascans themselves:

[There is a division in Malagasy between dialects with /d/ and dialects with /l/, which accounts for the coexistence of forms in Mal- and forms in Mad-.]

For the rest of it, it's just the typically quasi-regular patterning of derivational morphology, which happens to be especially erratic in the case of English adjectival forms of place names. (See e.g. "The evolutionary psychology of irregular morphology", 4/10/2008.)

Comments are open: insectivores welcome.


  1. Coby Lubliner said,

    February 7, 2009 @ 9:17 am

    When Madagascar first became independent and was known officially as République Malgache, rendered into English as Malagasy Republic, American media often referred to the country as "Malagasy (formerly Madagascar)".

  2. Cheryl Thornett said,

    February 7, 2009 @ 9:22 am

    I suspect that Geoff listened to the BBC Radio 4's Saturday Live programme last Saturday, which featured a Madagascan cockroach among other insects and spiders; in fact a Madagascan _hissing_ cockroach.

    Surely there should be a Language Log posting here somewhere?

  3. C.G. Armadas said,

    February 7, 2009 @ 9:33 am

    21 million is for "Madagascar" which google must have autoredirected you to. The real figure:
    "Results 1 – 55 of 55 for "madagascarene"

  4. marie-lucie said,

    February 7, 2009 @ 9:34 am

    I don't know Malagasy, but I saw a documentary made in Madagascar, which included a lot of talk, with subtitles, and as there were many names of places and people, I was surprised to see how much longer the written words were than the spoken ones: there seems to be a lot of unstressed vowel deletion (as with French e), hence English Malagasy = French malgache. The documentary was titled Mahaleo, the name of a musical group, pronounced ma(:)leo. Similarly, years ago the president's name was spelled Tsiranana, but on French radio we heard it as [tsiran].

  5. Adam said,

    February 7, 2009 @ 9:57 am

    I think Google does some odd synonym-replacement stuff in their searches, and I often fall foul of this sort of thing when I want to use Google for some slightly non-standard purposes (like this). Do you know how to turn it off?

  6. Tom Streetman said,

    February 7, 2009 @ 10:25 am

    Is that Johnny Depp in that picture? Perhaps pertaining to a new role? Willy Wonka: Bug Style!

  7. Mark Liberman said,

    February 7, 2009 @ 10:40 am

    Courtesy of Google Books, you can read David Griffith's 1854 Grammar of the Malagasy Language, which begins with this charming discussion:

    The Malagasy Language, abounding with vowels, is so mellifluous and soft, that it might be called the Italian of the Southern Hemisphere. Its character is so peculiar, philosophical, and original, as to render it truly amazing that uneducated, and semi-civilized people, should have preserved it in such perfection. They have no literature; the language has therefore reached its present state of excellence merely by ordinary conversation, speeches in the public assemblies, and pleadings in the courts of justice.

    With respect to marie-lucie's comment about Malagasy vowel deletion, we can note that the Italian of the northern hemisphere is also variably prone to vocalic dis-abundance.

  8. Tim McCormack said,

    February 7, 2009 @ 11:12 am

    @Adam: As far as I know, there's no setting that will turn off Google's new "sloppy search" with automatic "typo" correction. (Yay, derisive quotes!) However, you can enclose terms in double quotes and prepend a plus sign if you want to avoid it for any particular term.

  9. C.G. Armadas said,

    February 7, 2009 @ 11:22 am

    The growing "intelligence" of google is making it harder and harder to search for exact non-popular phrases and words.
    How I did it: I searched for "madagascarene".

    Google: "Did you mean: "madagascar"
    [I wish I could just click a NO button and have it disappear!]
    I scroll down to the section Results for: "madagascarene". Now to isolate this exact spelling I click on the Next button at the end of this section. Now it will display results 11-20 of Way too Many (it still provides the figure for its suggested phrase). In order to get a results number for the exact phrase you have to get to the last page of search results. Do this by clicking to a far page of results or manually editing the url. If you have surpassed the true number of results then it will show you the last page with the revised results figure.

    This search should give you the results for an exact word:
    Replace madagascarene with any word, and if there are less than 990 total results then it will give you the exact number of results. No more messing around with "helpful" suggestions.

    It would be so much easier if there were some new syntax for an exactly exact query.

  10. mollymooly said,

    February 7, 2009 @ 11:39 am

    "Madagasque" gets fewer ghits than I would like or have expected. Oh well.

    On a slightly related note, I've heard some American TV/movie characters pronounce the final syllable of "Madagascar" with a full vowel, like "car". How common is this? Dictionaries only list schwa, which is all I hear in real life.

  11. language hat said,

    February 7, 2009 @ 12:15 pm

    "Car"? Surely you mean "cad" (i.e., æ)? I pronounce it with a full /æ/ myself, and I don't think I've ever heard anyone use a schwa, which strikes me as an unlikely reduction in a syllable that by the laws of English morphophonemics carries a secondary stress.

  12. Karen said,

    February 7, 2009 @ 12:56 pm

    Well, hat, I use the same a in the last syllable of Madagascar as I do in "car", and it's not the one I use in "cad"… for what that's worth.

    Strikes me this is a bit like Canada/Canadian or Hawaii/Hawaiian – if Madagascar is a country (or bigger), you use the adjective; if it's a lesser entity you use the bare name – Tennessee walking horse, Kentucky bourbon, Florida oranges, American cheese, French wine, African swallow … Those dang places that start out regions and turn into countries get confusing (Canada geese, Bavarian chocolate, Hawaiian shirt…)

  13. mollymooly said,

    February 7, 2009 @ 1:28 pm

    Hat, did you misread "final syllable" as "first syllable"? If not, I don't understand your comment at all.

  14. Bob Ladd said,

    February 7, 2009 @ 2:43 pm

    Like mollymooly, I assume Language Hat misread the original comment. Like Karen, I have the full vowel of car in the final syllable of Madagascar, and in fact I'm really surprised to learn that the dictionary shows only schwa.

  15. Bob Ladd said,

    February 7, 2009 @ 2:46 pm

    And like C. G. Armadas, I wish Google would provide the option of replying "no" to one of their "Did you mean X?" queries. I had a very hard time finding something I was looking for on Google Scholar the other day precisely because Google was convinced I was looking for something else.

  16. Tim said,

    February 7, 2009 @ 2:55 pm

    If you follow the links in Wikipedia's source notes in their Madagascar article, it appears that Madagascar's embassy in the US claims that Malagasy is the correct term for the people and the language, and Madagascan is used by people who don't know better. However, they don't mention what term you should use for Madagascar-related concepts that are not people or languages.

    The National Geographic Society, on the other hand, says that Malagasy is only for the people and language of a specific ethnic group, and Madagascan is used for everything else.

  17. James Wimberley said,

    February 7, 2009 @ 5:28 pm

    A "Malagasy hissing cockroach" would be an impressive circus act, like an arithmetic tapping horse.

  18. Rubrick said,

    February 7, 2009 @ 6:13 pm

    That person is wearing way too much madagascara.

  19. Jongseong Park said,

    February 7, 2009 @ 8:44 pm

    My reaction was to wonder "Madagascan?"

    That was my first reaction as well.

    Marie-Lucie: I've also wondered about the pronunciation of Malagasy words being a lot shorter than the spellings indicate. The French name 'Tananarive' is apparently closer to the actual Malagasy pronunciation than the Malagasy spelling 'Antananarivo', which is the one used in English as well. Wikipedia's page on Malagasy mentions that unstressed vowels are devoiced and almost completely elided in many dialects, but I wonder if this is the whole story. Could anyone please shed light on this?

  20. Coby Lubliner said,

    February 7, 2009 @ 10:06 pm

    I have some anecdotal evidence about vowel elision in Malagasy. A number of years ago I heard the wonderful musical group Tarika introduced by its lead singer, the beautiful Hanitra. She presented the group as [tak] and herself as [hantʃ].

  21. dr pepper said,

    February 8, 2009 @ 12:18 am


    Unfortunately the article didn't say how many hisses they have for "lemur".

  22. Chas Belov said,

    February 8, 2009 @ 2:58 am

    The other problem with Google counts is that if you start going forward through them, you will usually eventually hit a notice that similar results have been removed. This often happens at a very low number of pages.

    I often try using quotes in the search. That is, I actually did type "madagascarene" with the quotes and it seems to have respected that, even though it does first offer me two results for "madagascar" first.

    But I see the complaint about the count. "Madagascarene" (with the quotes) initially shows a count of 114 million, but clicking on the link for results page 10 gives us the following message on page 6:

    In order to show you the most relevant results, we have omitted some entries very similar to the 51 already displayed. If you like, you can repeat the search with the omitted results included.

    If you use the option to repeat the search with these excluded results included, then click on the link for results page 10, we actually see a page of results where the the word "madagascarene" does appear. However, 8 of those results show a context of "The plan was adopted, the necessary treaty made, with legislation to carry out its provisions; the Madagascarene Philosopher took his seat in the Temple of …" meaning those really were duplicates that Google had previously removed.

    So to get a reliable count of an obscure (or even a non-obscure) word form, you have to:
    1) Type the desired word or phrase in quotes so the preferred spelling is not included in the search results.
    2) Navigate far enough into the search results such that Google eliminates the duplicates.

    Then you can find that the actual Google count for "madagascarene" is 51.

    And here's the code for a bookmarklet to do just that.

    javascript:q="+(window.getSelection?window.getSelection():document.getSelection?document.getSelection():document.selection.createRange().text);if(!q)q=prompt('Search words',");if(q!=null)location=''+escape(q).replace(/ /g,'+')+'%22&start=990';void 0;

    I'm assuming I can't post JavaScript as a link in your comment, so just create a new bookmark and make the above code the location for that bookmark. Adapted from

  23. Chas Belov said,

    February 8, 2009 @ 3:09 am

    Using the above bookmarklet, which I have called "ExactGoogle" on my Firefox toolbar, I get the following counts:

    Madagascan 680
    Madagascarene 51
    Madagascarian 425
    Madagascrian 11
    Malagasy 668

    Oddly, Google sometimes adjusts the count to the new, lower number, and sometimes doesn't. It would be interesting if someone could figure out the trigger.

    Hope this helps.

  24. Chas Belov said,

    February 8, 2009 @ 3:18 am

    Sorry for the multiple posts. It seems there are major limits to this tool which only Google can explain. The ExactGoogle count for Obama is 663 and for Cheney is 658, both of which are unrealistically low to say the least. So apparently there are other qualifications for relevance beyond duplicated contexts. Ah well, back to the drawing board.

  25. C.G. Armadas said,

    February 8, 2009 @ 3:55 am

    ExactGoogle only works to circumvent "suggested" entries, which usually are only included if the the search term isn't very popular, so it should be used after regular exact queries (quotation marks) searches are determined to have added suggested variations. Searches with more than 1,000 expected hits might have weird results.

    Oddly enough it seems to only produce a low result for "obama" Results 431 – 432 of 432 for "obama" whereas Results 651 – 658 of about 20,000,000 for "cheney"
    I won't begin to speculate on where the 432 or 658 numbers come from or why cheney shows the full number and not obama.

  26. language hat said,

    February 8, 2009 @ 10:20 am

    Hat, did you misread "final syllable" as "first syllable"? If not, I don't understand your comment at all.

    Yes, I did — mea culpa! But in the final syllable, I do have a schwa. I think that's a minority pronunciation, though.

  27. Skullturf Q. Beavispants said,

    February 8, 2009 @ 1:26 pm

    I say "Madagascar" with the last syllable rhyming with "car", as opposed to a schwa. I also don't say "Madagascar" very much at all. This could be an instance of the following rule: if you yourself don't say a place name very much, you'll tend to articulate it. If you say it a lot, you're more likely to elide certain sounds or turn vowels into schwas.

    I live in Toronto, and it's only people who aren't from here or don't have reason to talk about the place very much who say "Tore-On-Toe". We say "tuh-ronno".

  28. Chas Belov said,

    February 8, 2009 @ 10:21 pm

    @C.G. Armadas:

    I didn't find the suggested entries to be particularly intrusive, as there were only 2 of them. I think the big problem is that content is being duplicated so much throughout the Internet that the high number cannot be considered to be reliable.

    But if the low number is not reliable either (viz obama or cheney)–that is, both duplicated and non-duplicated content is being eliminated–then I'm not sure whether we can consider comparing word or phrase frequencies using Google to be useful. Can we reasonably assume that words common enough to be listed in the OED, such as Madagascrian, only occur 11 times on the searchable Internet? If I choose to remove the filter, it gets up to 19, with obvious duplicates. So where did the original search page 1 count of 15,200 (with no filter) come from?

    I don't think we can necessarily assume that all content is being duplicated equally, although it is certainly amenable to sample tests. But are Google word counts for entertainment purposes only?

  29. Matt Pearson said,

    February 9, 2009 @ 3:09 am

    In the Merina dialect of Malagasy, which I work on, there are four vowels, /a/, /e/, /i/, and /u/, where /u/ is spelled "o" and /i/ is spelled "i" except word-finally, where it's spelled "y". The high vowels /i/ and /u/ are devoiced (and usually very short) in word-medial and word-final unstressed syllables; while /a/ is devoiced in word-final unstressed syllables. /e/ is never devoiced. (Main stress normally falls on the next-to-last syllable, with secondary stress on every other syllable preceding, but there are some complications and exceptions to that rule–the major one is that words ending in "-tra", "-ka", or "-na" normally have main stress on the third-to-last syllable.)

    Also, the letter "h" is normally silent in Merina, so "Mahaleo" does come out /ma:leu/. As for "Tananarive" as the French version of "Antananarivo", the final vowel is indeed dropped from the French version in imitation of the devoicing of the final /u/ in the Malagasy version. However, the missing "an-" is not due to anything phonological; rather "an-" is a separate element, locative prefix found on many Malagasy place names, which for some reason got left off when the name was rendered in French. The name means "at the town of a thousand" ("an-" = "at", "tanana" = "village/town", "arivo" = "thousand").

    As for the use of the name "Malagasy", it's definitely used for both the language and the people. In Malagasy, the language is usually called "teny malagasy" (where "teny" means "word/language"), typically shortened to "teny gasy". The word "malagasy" or "gasy" is also used for cultural institutions associated with the people/language: for example, traditional folk songs, and the events at which they're performed, are called "hira gasy", where "hira" means "song". However, "Malagasy" is never used to refer to the island itself, which is called "Madagasikara" (with devoicing of the "i" and the final "a", of course). The official name of the country is "Repoblikan' i Madagasikara" ("Republic of Madagascar").

  30. language hat said,

    February 9, 2009 @ 9:19 am

    Thanks very much for that informative comment, Matt!

    Last night I was watching Bondarchuk's clunky but visually appealing movie of War and Peace, and was amused to be reminded that when poor heart-torn Natasha is trying to distract herself from her problems, she repeats the word "Madagascar" several times (in Russian, of course, with the stress on the final syllable and schwa in all the rest).

  31. Matt Pearson said,

    February 9, 2009 @ 4:32 pm

    The Russian pronunciation is pretty similar to the Malagasy pronunciation, then. In Malagasy it's more or less /, with main stress on /kja/ and devoicing/shortening indicated by parentheses. (It's /kja/, not /ka/, because /k/ and /g/ become palatalized after /i/, even when /i/ is devoiced.)

    Incidentally, I've always assumed that "Madagasikara" is derived from the European "Madagascar" (adjusted to Malagasy syllable structure), rather than the other way around. However, "Madagascar" itself is presumably derived from "Malagasy", with the /l/ ~ /d/ alternation that Mark mentions. I'm just not sure where the "-car" part comes from (there's no formative "kara" that I'm aware of in Malagasy place names).

  32. Jongseong Park said,

    February 10, 2009 @ 6:43 am

    Thanks for the informative comments, Matt Pearson. If you're still reading this, I hope you wouldn't mind me asking about how the sequences "tr" and "dr" are realized in the Merina dialect, especially when preceded by "n". Being Korean, I am interested in how to represent different languages most naturally in the Korean alphabet, in order to spell Malagasy proper nouns in Korean, for example.

  33. Mark Liberman said,

    February 10, 2009 @ 11:13 am

    Matt Pearson: I'm just not sure where the "-car" part comes from (there's no formative "kara" that I'm aware of in Malagasy place names).

    The OED says:

    < Middle French Madagascar (c1307 in Old French in form Madeigascar), of uncertain origin; perhaps < Arabic *Madaqas-barr < a Malagasy self-designation (compare Malagasy Madagasi MADAGASS n.) + Arabic barr land (compare *Madeigascar-bār ‘land of the Malagasy’ in Encycl. Islam (1984) s.v. Madagascar). Compare MALAGACHE n. and adj., MALAGASY adj. and n. The Malagasy name is Madagasikara but there is no Malagasy terminal element -(i)kara.

  34. Matt Pearson said,

    February 10, 2009 @ 2:31 pm

    OK, so I guess I'm not alone in not knowing where the "-car" comes from (it's good to know I'm not missing anything obvious). Thanks!

    As for "tr" and "dr" in Merina: I'm not entirely sure about these, to be perfectly honest, not being a phonetician. I always pronounce them as apico-alveolar affricates (probably with some retroflexion), and no native speaker has ever corrected me, so I suppose I'm reasonably close ("ntr" and "ndr" are just prenasalized versions of "tr" and "dr"; each of these four should probably be considered a single segment in Malagasy).

RSS feed for comments on this post