My Hovercraft is Full of Eels
or, Why I Love Omniglot

Before we Begin

* If you enjoy this kind of thing, there is a similar list here, devoted to the even more useful statement “I can eat glass without hurting myself”. Unfor­tunately the Inuktitut version is specta­cularly un-idiomatic and only marginally gramma­tical. I know this for a fact, because I contri­buted it myself.

It was three, but I followed their link.

Where else but omniglot would you find a whole page of translations of this inarguably useful utterance?* For the benefit of the two people who don’t already know it, they’ve even got a link to the original Monty Python sketch.

If you are in a hurry and blundered onto this page by mistake: The phrase appears to mean “Please give me a box of matches.” But don’t take my word for it; see for yourself.

This is my favorite:

ᐅᒥᐊᕐᔫᑉ ᐳᓪᓕᓕᒫᐸᒐ ᑕᑦᑕᕐᓂᖅ ᐊᒻᒪᔭᖅ
Umiaryuap Publimaaqpaga tattaurniq ammayaq.
Umiarjuap Pullimmaqpaga tattaurniq nimiraq.

I bet you think I am making this up. But in fact it’s even better, because in 2010 I did not know that in Dialect 2 of Language B, “yes” is not ii but aa.

I do have a few problems with this. The overarching problem is that when you muck about in linguistics to any extent, you lose the ability to say “I do not know this language.” At most, under extreme duress, you might admit that you only know three words and the rule for forming plurals. This can be a problem when the only thing you know about Language A is that “no” is iie, and the only thing you know about Language B is that “yes” is ii. Or, conversely, that in Language C acchaa is “OK”, while back in Language B, aakka is “no”. Sooner or later this will get you into trouble.

So is this two dialects or three? Never mind y vs. j; that’s just ortho­graphy. The choice between -bl- and -ll- is also a perfectly legitimate dialectal variation. It’s the same thing that turns iglu into illu. Problem is, if you say “publi­maaqpaga” you also have to say “kabloona”. Which, to this kabloona, sounds more like something that would come out of a Klondike prospector’s mouth.

On the other hand, consonant clusters are useful when you’re learning the language. It does not take a lot of linguistic sophistication to convert -bl- and -gl- into -ll- on the fly. But if all you know is -ll-, you don’t know the underlying form. It might even be -dl-. So let’s make it publ-, written ᐳᑉᓕ- (pupl-) because the syllabary has no separate letter b. It doesn’t need one, because the sound only occurs in environments where p is impossible.

The syllabary doesn’t match the first Roman line. Is it meant to? It’s obviously not the same as the second line, which uses a different word for “eel”. Trans­literated, we’ve got Umiarjuup pullili­maapaga tattarniq ammajaq. The eel is fine; we’ll get back to the other three words later.

Last call

About half a year after my entanglement with eels and hovercraft first started, I went back and took a closer look at this page. There are passages where I have no idea what I was talking about. Truth.

Rather than go back and rewrite the whole thing, as I did with duct tape, I offer this oppor­tunity to jump over to:

That was your final chance. The hovercraft is now boarding.

rats going for a cruise in a hovercraft

Gentlemen, Start Your Engines

I don’t think it says “full of eels” (descriptive phrase) at all. I think it says “filled with or by eels” (passive verb with explicit subject and object, or agent and patient if you prefer). This is bad, because Inuktitut has a nasty habit of dragging in every word that has even the remotest connection to the action, and making all of them get involved in your choice of endings. It is probably the nouns’ revenge because there are a piddling eight noun endings as against hundreds of verb endings.

* Proverb: Even Franz Boas nods.

Unlike the business about snow,* this is literally true. Or at least half true. Discrete morphemes, every one of ’em. The official count is—hang on while I look this up—“699 verb endings in the North Baffin dialect . . . 609 in South Baffin”. Clearly South Baffin is the place to go if memor­izing endings makes you tired. Not a thing you can do about it, either. You could dump all the other parts of speech, but throw out the verbs and there would be no Inuktitut.

It’s the first half of the contention—the part about the eight noun endings—that’s got me scratching my head. What are all those hundreds of possessive forms, if not noun endings?

Yup, I said hundreds. For starters, you’ve got four possible grammatical persons doing the owning, because obviously Inuktitut is not going to let you waffle about whether something belongs to the person you’re talking about (third person) or to some other guy (fourth person). The other guy might object too. Then there are three numbers for the owner, and three more numbers for the thing owned, for a total of thirty-six. And then you multiply those by the eight possible “cases” (in quotation marks because “case” is really an inflected-language concept, while here we’re talking about affixes). Grand total—don’t take my word for it, go get your own calculator—288.

* If Boas can make silly blunders, why should I be any different? The full expla­nation will have to wait until we start eating glass. For the time being, let’s just stipulate that whenever I say Causative, I really mean, er, Becau­sative. That is, “Because such-and-such”, not to be confused with “Causing someone or something to do or be such-and-such”.

Wait. It gets worse. Unlike verbs, the possessive affixes don’t fall into a nice pattern of shared forms and consistent stems. You’ve got to memorize them all by brute force. And, also unlike the verbs, they don’t simply have endings. They’ve got morpho­phonemic rules. With verbs, you rarely have to remember more than that there is a vowel-stem form and a consonant-stem form: -jara versus -tara, say. And maybe an additional twitch if your stem ends in q or t. (The letter t tends to be fragile, and q can be cantankerous.) Or if it’s a Conditional or Causative*, because voiceless stops don’t get along well with g. Oh, and the Frequen­tative can get downright nasty.

All of this pales by comparison with the nouns. With nouns, not all affixes are alike. Some final k’s and q’s get deleted. Some pair up with the beginning of the affix: k plus tt reduces to kt. Some assimi­late: k plus n turns into ngn—or, if you prefer, ŋn. We will not even talk about nouns ending in t. The subject is too painful.

Finally, let’s not overlook the 24 (eight “cases” × three numbers) plain ol’ non-possessive forms. Aggregate total: 312 servings of, I guess, chopped liver.

* The syllables -ga- and -gu- respectively. And, just to keep you on your toes, there’s a further quirk whereby a Causative ending in -gami- corre­sponds to a Condi­tional ending in -guni-. And the fourth person insists on going its own way, giving you m and p instead of ga and gu. But it’s still better than memorizing 66 separate endings.

If you’re really a glutton for punishment, you could also point out that, for example, there isn’t really one vast set of Causative endings and a whole other vast set of Conditional endings. Look more closely and you’ll find a Causative doodad and a Condi­tional doodad,* each followed by the exact same endings. So if you know one form, you automat­ically know the other. So, bit by bit, the Balance of Affixes shifts from the verb side to the noun side. But something tells me that if I disagree with Mick Mallon, I will end up being very, very sorry.

ᐅᒥᐊᕐᔫᑉ, Umiaryuap or Umiarjuap

I bet you never realized how many Inuktitut words you already know. Peel away the affixes and you’re left with ᐅᒥᐊᖅ, umiaq. Boat. You knew that one. Put the main affix back and you’ve got umiaq + juaq, a big boat. We’ve got a lot of eels to transport, so your ordinary family-sized umiaq isn’t going to cut it.

That q at the end of the boat isn’t going to cut it, either. Voiced consonants such as j don’t like getting too close to unvoiced ones. And, since the three consonants that can come at the end of an Inuktitut word are all unvoiced, this keeps the j busy. Especially the one in -juaq, because it’s got a different rule for every one of the letters it might meet—one for each consonant, another for the vowels collectively. Those rules are, of course, for the other letters, not the j itself. Most affixes are not interested in compromise. I got here last, so I get to make the rules.*

* Ethnopolitical analogies present them­selves, but let’s not go there.

In this case, the j folds its arms, sits back and waits for the q to turn into an r, because the letter r is considered the voiced form of q. And the nasal of q. Don’t look at me. I didn’t invent the phonetic structure of the language. If it makes you feel any better, it is not considered, well, an r. So you never have to pay it triple overtime.

Right about now, someone is objecting that it doesn’t say umiarjuaq but umiarjuap with an abso­lutely unambiguous p at the end. Well, yes. I would be happier if it were q. But that’s because I was not brought up speaking an ergative language and it doesn’t come naturally to me. Besides, I’m not going to argue with three matching p (for -up) endings. If it looks odd, it should: we have run into the only suffix in the entire language that ends in p. It has only one job, and it does it well: mark the subject of certain types of verb. About which, more later.

* If you ever hear three consonants in succession, run like hell, because there is a nanuq right behind you.

I’m fudging this slightly. The first u is attached to a conso­nant, so really you would only have ᔪᐊᐅᑉ. Stop quibbling, willya.

I hate going with the Roman over the syllabary, but unless they’re in some truly arcane dialect, there’s no way to justify the -uup. First, the -up suffix eats the final q of the main word. Which you might say serves -juaq right for forcing that earlier q to change. Since this leaves us with three conse­cutive vowels—which is almost as bad as three conse­cutive consonants*—the initial u of the suffix also has to go. No chance of ᐅᐊᐅᑉ (uaup), no matter how pictur­esque it looks. You might think it would be easier to stick with umiar­juaqup, leaving both main word and suffix plainly visible to the naked eye. But we are bound by iron­clad morpho­phonemic rules that admit of no argument or discussion.

This is assuming that we are talking about some specific and particular group of eels, because if you really want to split hairs, the distinction in Inuktitut is not between transitive and intran­sitive. It’s between “specific” and “non-specific”. Do you need to see a man about a dog, or do you have some particular, individual breeder in mind? Are you on first-name terms with the eels? If not, the whole -up business flies out the window.

ᐳᓪᓕᓕᒫᐸᒐ, Publimaaqpaga or Pullimmaqpaga

First, a word about the word. All three versions of it. Roman publimaaqpaga and pullimmaqpaga, syllabic pullilimaapaga. For dialectal convenience we’ll level them out to publimaaqpaga, publimmaqpaga, and publilimaapaga. The language won’t mind. When your dialectal continuum extends from the northern fringes of Alaska to the east coast of Greenland, you learn to be accommodating.

That extra li in the syllabic version looks at first sight like an obvious typo. But there’s no law that says you can’t stick an affix starting with l onto a word whose last syllable also starts with l—and there happens to be just such an affix, -limaaq, meaning all or the whole thing. We don’t have air pressure on the left balanced by a paddle wheel on the right; we’re hovering all the way around.

The part that may really be a typo is the first vowel of the pair. Unless we’re in dialectal-variation territory again, it practically has to be an a. That’s publaq, meaning air bubble. A very large one in this case, to go with our very large umiaq. Now, I know about as much about hovercraft as I do about the Inuktitut language, so I’m not going to speculate about the pa or possibly paa. There are several available suffixes, including one that just makes the whole thing bigger. Besides, we are talking about a Lexicalized Composite Root, meaning that you don’t actually have to pick it apart.

In any case, I am perfectly happy with the -ga at the end. That’s “my” when you’ve only got one of some­thing. Not to be confused with -kkak if you’ve got a matched pair of hover­craft, or -kka if you’ve got a whole fleet. But my happiness may be premature. We’re assuming the base form ends in a vowel: publa­limaaqpaa, while I’ve got a nasty feeling it’s really publa­limaaqpaaq. That changes the last syllable into -ra, due to morpho­phonemic rule that kicks in when certain suffixes in g are attached to a word ending in q. Not all suffixes in g, of course. That would be too easy.

I would really like to take the translators’ word for it that only the hover (the publalimaaqpaa or possibly publa­limaaqpaaq, when it’s just sitting there not doing anything) and not the craft (umiarjuaq, ditto) belongs to me. And, conversely, that only the craft and not the hover gets the -up. On some level it does make sense. It would be just as easy to fill it with eels if it were an ordinary ship. Easier, possibly.

This is not merely about respect for the translators’ expertise. The alternative is—you know what’s coming, don’t you?—to dig up a single affix that conveys both pieces of information. The thing belongs to me and it’s performing the grammatical -up function. That would give us the affix -ma . . . unless, of course, we are attaching it to a word ending in q. Putting a nasal next to a stop is just as bad as putting voiced and voiceless consonants side by side. The q has to change to r—an exercise that should be very familiar to it by now.

ᑕᑦᑕᕐᓂᖅ or tattaurniq

Changing from q to r is, by amazing coincidence, precisely what happened in tataurniq, our verb. It is made up of the three pieces tatat-, -uq- and -niq: by Inuktitut standards, barely enough to permit it to show its face in public.

* In the same sense that Alonzo and Melissa ranks among the highest achieve­ments of American litera­ture, or Coming Home among text-adventure games.

More precisely: die staunens­werte Zusammen­setzungs­fähigkeit, die die Nenn- und Zeitwörter haben. So, in fairness, it’s not the whole language. Just the nouns and verbs.

Digression: Scholars of Inuktitut do not agree on a lot, but nobody can deny that the language has long words. Call it a universal postulate, like “What Ottawa doesn’t know won’t hurt them.” The Reverend Edmund Peck’s Eskimo Grammar, which surely ranks among the highest achievements of missionary linguistics*, says that “a word may be formed of a length truly astonishing”. On the other side of the linguistic and religious fence, Theodor Bourquin warns the student—with a perfectly straight face—that a charac­teristic feature of the language is its staunens­werte Zusammen­setzungs­fähigkeit (emphasis his). Two minds with but a single thought.

To illustrate the point, he offers up this tasty two-word utterance.


Or, if you prefer:


Labradorian uses e and o though they’re not really phonemic; the Moravians invented the letter kra (ĸ) for the sound now written q; and dl corresponds to the Baffin ł. But with all those diacritics, I’m not going to hazard a guess which vowels are supposed to be long.

* Oh, all right. “But even if we have not always been able to feel fully comfortable, O let us yet ever struggle to be very thankful to him!”

For a given definition of “comfort­able”, anyway.

Either way, it translates—according to Bourquin—as simply

Aber wenn wir uns auch nicht immer haben völlig wohl befinden können,
o so lasst uns doch uns bestreben, ihm immer sehr dankbar zu sein!

You can see his point. On the one hand: clear, lucid, intelligible, 1885-vintage German. On the other hand: Inuktitut.

Well, drat. I was hoping that when I came back the affixes would have sorted themselves out: tatat-, something, something. But no luck. The Morphological Analyzer insists that -uq- is a frequen­tative used when both subject and object are plural. “When my fleet of hover­craft is filled with eels, I habitually . . .” Uh . . . Nope, we’re missing something here. (When the tool is good, it is very very good. But when it goes astray, watch out.) Problem is, I can’t find anything else for it to mean. At best, it might mean that the filling started at some time in the past and is still in progress as we speak. Or that the eels have not yet found a way to escape. That doesn’t get us far.

* This does not mean exactly what it sounds like, but it’s pretty close. A copula is a linking verb, like “is” or “were”. Authors of dead-language textbooks are fond of hitting you with things like καλοὶ οἱ ἐν τῇ σκηνῇ στρατιῶται (the soldiers in the tent [are] handsome) in order to postpone the moment when they have to start teaching verbs. This is why nobody has a solid grip on the Aorist Passive in Greek, in spite of its measly total of six forms. Maybe eight if you want to go all Homeric and include the dual.

And -niq is worse. Sure, there are other -niq- affixes, but nothing that can come word-finally. As it is, it’s a verb-to-noun converter, making the whole package mean something like “the state or act of being habitually filled”. Not necessarily with eels, of course. But unless you perform some jiggery-pokery elsewhere in the sentence, you are now in trouble, because you’ve wiped out your only verb. Some languages are happy to do this, falling back on tricks like the Implied Copula*, but don’t even think of trying it here.

So let’s keep it as a verb. The question is: which one? There are two tatat- verbs. Or, at least, one and a half. The main one behaves like a lot of Inuktitut verbs: its meaning turns around depending on which type of ending you use. With specific or “transitive” endings, it’s a transitive verb meaning to fill something. If you are standing there pouring eels from a bucket into your hovercraft, you get control of the verb—and your very own -up ending, if you choose to include yourself by name—while the hovercraft sits there in un-affixed glory.

If, on the other hand, you are just scattering eels in a general sort of way, and some of them happen to land in your hovercraft, the whole sentence flip-flops. You lose your affix; the verb changes to a non-specific or “intransitive” ending, with preceding -si- to keep it functionally transitive; and the hovercraft—assuming you bother to mention it at all—picks up a -mik. Or possibly -nnik because, with or without eels, it definitely belongs to you.

* Inuktitut has as many descrip­tive words as anyone, but gramma­tically they’re not adjectives. They’re . . . drumroll . . . verbs. Ordinary intransitive verbs, conju­gated just like any other verb. They just happen to mean “to be red”, “to be tired”, ”to be full of eels”, and so on.

And if you leave yourself out of the sentence in order to concentrate on the hovercraft, the verb gets a simple non-specific ending and its meaning switches from “fill” to “be filled”. This last form is probably indistin­guishable to the naked eye from tatat- no. 2, an adjective-verb* meaning to be full.

ᐊᒻᒪᔭᖅ, ammayaq or nimiraq

It’s time to take a closer look at that eel. Singular. It’s sitting there stark naked at the end of the sentence, devoid of affixes. This is all well and good if it’s the patient of the verb—the same gramma­tical doodah that resulted in the -p at the end of the hover­craft. Except that I kinda think, in that case, it’s supposed to come before the verb. Ergative languages are appar­ently not big on SVO word order.

* The answer, astoundingly, appears to be no. Missed an opportunity there, Inuktitut.

Wherever it is, it must be a very large eel, because by any name—ᐊᒻᒪᔭᖅ (ammajaq) or nimiraq—there’s only one of it. If there were more of them, would we have to go back and change the -up into something else?* Is this some kind of collective, abstract, conceptual eelness, or did the trans­lator(s) simply get tired at this point, throw in an eel and say I’m done?

* Short version: “subject” and “object” are gramma­tical terms, while “agent” and “patient” are real-life terms. The long version is, oh, I dunno, around here some­where.

It would be nice if we could make -niq into a verb ending. If so, it would tell us both how many agents (“subjects”) and how many patients (“objects”)* are involved, leaving no need to repeat the infor­mation. Unlike inflected languages, Inuktitut does not demand long strings of matching endings. None of this exhausting and redundant business with my (nominative singular feminine possessive adjective) craft (nominative singular feminine noun) hover (nominative singular feminine adjective) that makes Latin and German so much fun. The -ga tells you how many there are—just one, in this case—and who it belongs to. The -niq tells you how many things are acting on how many other things. Except that, well, it doesn’t.

There’s a further problem. We’ve been talking as if the eels are either the patients or the agents of the verb. They may have been the patients of some earlier verb that caused them to end up filling your hovercraft, but nobody is doing anything to them now. And I don’t think anyone is prepared to suggest that the eels, of their own free volition, threw themselves into the hovercraft. There are situations where “My hovercraft is filled with (or by) eels” is just another way of saying “Eels fill my hovercraft”, but this is not one of them. Gramma­tically they’re on the sidelines, even if physically they are sloshing all over the deck of your hovercraft and making it impossible for you to think of anything else.

Happily, there is an alternative approach that the translators don’t seem to have tried. You may not need that tatat- verb at all. Take your eels—plural—stick on the appro­priate affix to change them into a verb meaning “is full of . . .”, add a verb ending and you’re done. My hovercraft is now duly eel-filled.


Earlier in the evolution of this page, I added a fourth and fifth dialect to make it

ᐅᒥᐊᕐᔪᐊᑉ ᐳᑉᓚᓕᒫᖅᐹᒐ ᐊᒻᒪᔭᖅ ᑕᑕᐅᕐᓂᖅ
Umiarjuap publalimaaqpaaga ammajaq tataurniq


ᐅᒥᐊᕐᔪᐊᖅ ᐳᑉᓚᓕᒫᖅᐹᒐ ᐊᒻᒪᔭᕐᓂᑦ ᑕᑕᑦᑐᖅ
Umiarjuaq publalimaaqpaaga ammajarnit tatattuq.

Having no idea what to do with the eels, I picked an ending at random. The odds were one in seven that I’d guessed right. When your hovercraft is full of eels, you’ll take any odds you can get.

Make that one in . . . uhm . . . six and a half. Doesn’t it seem reasonable that the agent of a passive verb would be ablative? Sure— if you live at the Kalaallisut end of the dialect continuum. Any further west and it becomes decisively dative: not ᒥᑦ (-mit) but ᒧᑦ (-mut). Or ᓂᑦ (-nit) and ᓄᑦ (-nut), depending on number of eels.

As I continue my quest for just the right affix, let’s move along to Dialect Six. Among other other things, I’m going to take possession of the entire hovercraft—and get those eels into the proper case while I’m at it.

* Answers beginning “the first syllable of the first word” will not make me happy. But then, who ever said the Inuktitut language was put on this earth to make me happy?

If you need some pointers on how not to say it, I’ve got plenty of those.

ᐅᒥᐊᕐᔪᐊᕋ ᐳᑉᓚᓕᒫᖅᐹᒐ ᐊᒻᒪᔭᕐᓄᑦ ᑕᑕᑦᑐᖅ
Umiarjuara publalimaaqpaaga ammajarnut tatattuq.

Someone will be along shortly to tell me where I went wrong.* Anyone know how to say “If it ain’t broke, don’t fix it” in Inuktitut?

The ^Second to Last Word

I have made a mournful discovery. Or rather, I have failed to make the expected happy discovery. There does not seem to be an affix meaning “is full of”. Have an eel, go hunting for eels, capture an eel, acquire an eel, catch sight of an eel: check. Have a big eel, have lots of eels: check. Have no eels, sell (or give away or lose) an eel, have almost no more eels, run out of eels, have inferior eels: check. Sound like an eel, smell or taste like an eel, be too close to an eel: check.

Be full of eels: no luck. It would seem to defy all probability that there exists an affix that Inuktitut does not have. But if it is there, it is hiding very thoroughly.

* Or, if you prefer, बहुव्रीहि. Do not try modernizing it to bahut chaaval; it won’t work.

Granted, some of those affixes might be made to work. But there’s the further risk that they are meant to apply specifically to people. Or to sea mammals, or snow­drifts. You can’t go around calling a squirrel a bahuvrihi* just because its nest is well-stocked.

Over on the noun-to-noun side we’ve got the predictable array of places or containers for eels. But that makes it sound as if I’m using my hovercraft as an expensive freight vessel or storage facility. There are also assorted people who either possess eels or, for variety’s sake, are bereft of eels.

The one that did catch my fancy was -ᑑᖅ- (-tuuq-), meaning abounding in eels. The abundance is built in; all you need is one eel and the affix will do the rest. As a bonus, the affix doesn’t do a thing to the preceding word, so the eels need have no worries about someone coming along and biting off their tails. This is especially reassuring when you’ve got a singular eel standing in for the entire group: bite one tail and you’ve bitten them all.

Normally, -tuuq- applies to game animals. If you are hunting caribou, it is obviously worth noting when you have found a place with lots of them. Can anyone doubt that my hovercraft abounds in eels?

There’s one remaining problem: having ditched the tatat-, we are now verbless. But for this problem there is an affix, the simple letter -u-. Just slap it on and your noun magically changes into a verb, ready and willing to accept any inflectional ending you choose to apply. Now, it happens that -u-, like so many affixes, eats the preceding consonant, leaving us with the dreaded Three-Vowel Sequence. The -tuuq- gets no credit for respecting the eels. But a one-letter affix in a language rich in—one might even say, abounding in—final consonants can be relied upon to know what to do. In fact that’s a general truth about Inuktitut affixes: they always clean up their own phonetic messes. In this case, the fix is to shove in an ng (one letter). That’s enough to break up the vowels and keep them from getting into mischief.

Dialect Seven:

ᐅᒥᐊᕐᔪᐊᕋ ᐳᑉᓚᓕᒫᖅᐹᒐ ᐊᒻᒪᔭᖅᑑᖑᔪᖅ
Umiarjuara publalimaaqpaaga ammajaqtuungujuq.

Bring on the harpoons. Eel season is now open in my hovercraft.