LINGUIST List 11.2149

Thu Oct 5 2000

Sum: Indeclinable Words

Editor for this issue: Lydia Grebenyova <>


  1. Mike Maxwell, Indeclinable Words

Message 1: Indeclinable Words

Date: Tue, 3 Oct 2000 17:52:43 -0400
From: Mike Maxwell <>
Subject: Indeclinable Words

In Linguist List 11.1948, I posted a query concerning "indeclinable words."
Herewith the promised summary. It's a bit longer than most summaries, I
guess because I feel it's necessary to include the examples people gave
(otherwise the claims can be hard to understand, at least to me).

First, I should say that several people questioned my use of the term
"indeclinable", since "declension" is usually used for the paradigm of nouns
and adjectives, and "conjugation" for the paradigm of verbs. I simply used
the traditional word "indeclinable", which presumably is used because this
phenomenon rarely happens with verbs. I know of no reason this could not
happen with verbs, or with other categories which are conjugated, so perhaps
the word "uninflectable" would be more appropriate. Neither "indeclinable"
nor "uninflectable" exactly rolls off the tongue...

In my original query, I mentioned several situations where words often
do not take inflection. One is of course categories which, in languages
otherwise ripe with inflection, are uninflected: conjunctions, prepositions
and adverbs often behave this way. (There are languages where one or
the other of these categories can be inflected--Irish has inflecting
prepositions, as Chris Cleirigh reminded me.)

More interesting (to me) is the situation where a subset of the words of
some category do not take inflection. I mentioned in my original message
English modal verbs, which are uninflected for person. (They also don't
have participial forms, but that might be for other reasons.) One might
attribute their behavior in this case to their belonging to a particular
subcategory of verbs. (One might also claim that that is a description
of the problem, not a solution.) A more common case (pardon the pun)
where a subset of the words of a category may be indeclinable is loan

My question was whether there were other instances of indeclinable words
that did not fall into either of these two classes (subcategories or loan
words). I offered one possible example: the Latin cardinal numerals from
quattour to centum (and mille). I also asked whether indeclinable words
could take derivational affixes, and if so, what the status of inflection on
those derived words was.

The respondents: Larisa Zlatic (, Stan Whitley
(, Betsy McCall (, Ivan
Derzhanski (, Martin Haspelmath (, Kim
Ebensgaard Jensen (, Bill Idsardi
(, Gladney Frank (, and Nancy
Hall (

Larisa Zlatic wrote back with details on Serbian indeclinable words:
> female names that do not end in ?a (e.g. Lori, Kejt,
>Suzan) and cardinal numerals 5 and above (exception is
>'hundred', 'thousand'). These two types of nouns have a very
>similar (restricted) distribution within a sentence.

(She and SteveWechsler wrote a paper on this topic: "Syntax and
morphological realization in Serbian", published in "Slavic in HPSG", 1999,
edited by R. Borsley and A. Przepiorkowski, CSLI, Stanford.)

Larisa further wrote:
>The indeclinable female names can take derivational suffixes, for
>example, the possessive suffix ?in, as in: Suzan-in 'Susan's', or
>Kejt-in 'Kate's'. Once they take that suffix, such words become
>'declinable, i.e., inflectional affixes can attach to it, as in:
>Suzan-in-om 'Susan's-instrumental, sg.' (meaning 'with Susan's)
>Kejt-in-og "Kate's ? genitive, masculine, sg.' (meaning 'of Kate's')
(I just noticed in editing this reply the '?'s here--I suspect it is a
character that my email has mangled, not the glottal stop.)

Stan Whitley wrote:
> Spanish... the syntactic use of nouns as color adjectives
>without taking on adjective morphology: unos pantalones cafe'
>/ (color de) vino / rosa/ / violeta /naranja ... Some native
>speakers do adapt some of them to adjectival agreement
>patterns but in standard Spanish they stay uninflected and
>no one (to the best of my knowledge) would add other
>adjective endings (e.g. -i'simo), much less the derivational
>morphology of true adjectives (*la cafeidad de los pantalones,
>*mis pantalones se encafecieron...)

I'm trying to learn French now, and this reminded me that I have been taught
that French color adjectives based on nouns ('orange', for example) are also
uninflectable. (I don't know whether they can take any derivational
affixes; perhaps someone could comment?)

Stan also wrote:
>...likewise Spanish "un millo'n (de)", which has to be a noun
>leading off a partitive phrase, unassimilated to normal number
>positions within the NP.

Here I think the reference is to the syntax, not the morphology: it's
'dos personas' "two persons", but "dos millones _de_ personas"
"two millions of (lit.) persons". But in 'dos mil personas', "two
thousand persons", the Spanish 'mil' "thousand" doesn't get
pluralized, unlike 'cien/ cientos' "hundred(s)" or 'dozena(s)'
"dozen(s)"--is this a hangover all the way from the Latin
indeclinable 'mille'?

And speaking of Latin, Betsy McCall wrote
>...the numbers from five to 100 are indeclinable in other old
>Indo-European languages, as well. Ancient Greek in particular
>shares this feature. Is this a mutually inherited feature of Proto-
>Indo-European? I don't know. Another "number" word which
>doesn't decline in Latin is "nothing" (though the equivalent oudeis
>does in Greek).

I would have guessed that indeclinability was a rather unstable thing, but
apparently it can hang on for millenia. Why should it be so stable? Sounds
like there's material for a thesis in that...

OTOH, the facts about color adjectives which "are" nouns (synchronically? or
only diachronically??) in both Spanish and French cannot (I presume) be a
"hangover" from Latin. How is it that the same construction has arisen in
both languages? (And what about other Romance languages?) At least in the
case of "naranja/orange", the noun would be pluralizable: "El ha comido dos
naranjas." Is either "naranja" or "orange" in their adjectival use
pluralized to agree with a plural noun (as opposed to agreement wrt gender)?
If not, why not? Neither language, to my knowledge, has productive
noun-noun compounding; what kind of construction is "una camisa
naranja", and where did it come from?

Ivan Derzhanski mentioned that acronyms are indeclinable in Russian
(and I would guess in many other languages, too). He also wrote:
>Arabic has a class of nouns that have the same form for all cases
>because they end in a vowel which absorbs the case ending
>(<dunyA> `world', <_dikr_A> `remembrance'). These can
>be pluralised though, and dualised too, so they aren't exactly
>what you are looking for, but they may be a pointer.

>In Spanish, where the only inflexion nouns can get is the plural
>_-s_, penultimately stressed _-s_-final nouns don't get it:
>sg. _la crisis_, pl. _las crisis_. Again the reason is phonological

Indeed, phonology is doubtless a common source of indeclinability. Perhaps
English passed through a stage like that on its way to losing case
And of course there are English nouns which don't take overt possessive
clitics for many speakers, namely those ending in schwa + s: Jesus, 
Pegasus, etc., not to mention plurals. Further examples of words which are 
uninflectable for phonological reasons include the language--

>...Maybrat (Bird's Head, West Papua): verbs with two consonants
>at the beginning don't take any inflection, cf.: ait y-amo 'he goes',
>nuo n-amo 'you go', but ait frok 'he emerges', nuo frok 'you emerge.
>(The data is from Philomena Dol's 1999 Leiden dissertation)

The above quote is from Martin Haspelmath, who also wrote:

>Masculine person names are indeclinable in Romanian --
>when they occur in a genitive position, the pronoun 'he'
>must be used in apposition to them.

An odd constraint, that. Along the same lines, Gladney Frank wrote:

>Also in Russian, a woman's name ending in a consonant is not
>declined. But this includes many names you'd hesitate to call
>foreign, e.g., _Borodich_.

Martin Haspelmath also mentioned "Ursula Doleschal
(Vienna,, who recently
completed a comprehensive study of indeclinability in Slavic."
Martin calls her "the world expert on this topic." A lead I'll want
to follow up.

I just remembered that there is a semantic class of nouns in English which
does not take overt plurals (besides mass nouns): deer, bison, antelope...
Jim Hoard once pointed out to me that this is the class of game animals (as
opposed to domesticated animals or wild non-game animals). I can think
of no good reason for this; when did it originate? Middle English??

Ivan Derzhanski also wrote, with regard to derivational morphology on
uninflectable words:

>...vowel-final stem is a major hurdle for Slavic morphology,
>derivational as well as inflexional. Otoh, feminine nouns with
>stems ending in hard consonants don't inflect in Russian, but
>they are legitimate derivation material.

OTOH, there are loanwords in Russian which cannot take derivational suffixes
(further stretching my use of "indeclinable"!):

>..._kashne_ and _pensne_ (from French _cache-nez_ and
>_pince-nez_). Or _interv'ju_. Or _ehmu_ (the stem is too
>short to lose the final _u_, so a chick emu would hardly be
>called *_ehm�nok_).

Further with regard to Russian, several respondents mentioned 'taksi'
"taxi" and 'kenguru' "kangaroo" as being uninflectable loanwords. But
their derivatives, 'taksist' "taxi driver" and 'kengur�nok' "kangaroo" do
take inflection:

>The genitive/accusative is _taksista_ and _kengur�nka_, etc
>_pal'to_ `overcoat': _pal'tishko_ `wee overcoat' (affectionate
>or deprecative). From _kofe_ `coffee': _kofejnyj_ `(of) coffee'
>(derived adjective). [all these derivatives take inflection--MM]
>Generally _-ist_ seems to attach to _-i_-final stems, merging its
>vowel with theirs (I've even heard _sist_ `C programmer, C fan').
>They don't normally get any other affixes though. On the whole
>stems in _-i_ and _-u_ are more inert than stems in _-a_ and _-o_
>(not quite sure where _-e_ belongs), and the shorter a stem is,
>the less likely it is to give up its final vowel just so that
>something can be derived from it.
>People have attached suffixes to a vowel-final stem under duress,
>esp. to Chinese names (_ciskij knjaz'_ `Qi prince'), but this looks
>dreadfully artificial to me.

(the above quoted from Ivan, with similar comments from Gladney Frank)

...and Bill Idsardi writes:

>Russian, loan nouns will be indeclinable only if they don't
>look like they belong to any declension class. For example,
>'pizza' is declinable, belonging to the -a (fem.) class
>(pizza, pizzu, ...). However, 'menu' is indeclinable.
>Presumably this is due to the fact that there is no -u
>declension (in terms of a citation form -- the -a declension
>has an -u member, the accusative). So I think the question
>is best thought about as a memorization problem for
>the learner -- when hearing such a word, does it
>sound like it belongs to an existing declension class?

The comment there about _citation forms_ caught my eye. When a Russian
speaker hears 'menu', why doesn't he just assume it's in the accusative? (I
realize that in the context of a sentence, the case is often identifiable by
other means. But certainly it can't always be the case (pun, again) that
borrowed words first appear in nominative positions.)

Along related lines, Gladney Frank writes:

>Consider Polish nouns borrowed from Latin with _-um_,
>like _muzeum_ and _centrum_: in the plural they lose the
>_-um_ and decline normally: _muzea_, _muze'ow_,
>_muzeach_, etc. Makes me wonder, is it indeclinable
>in the singular, or declined with the "singular" ending _

Kim Ebensgaard Jensen writes:

>Danish has lots of indeclinable words. One example
>is "ihjel" which means, roughly, "to death". "Ihjel" is
>indeclinable, and what is also interesting is that it never
>appears alone - it always occurs in connection with a
>verb in more ore less idiomatic constructions as in
>"sl� ihjel" (kill), "arbejde sig ihjel" (work oneself to death),
>and "keden nogen ihjel" (bore somebody to death).
>If you take a look at the VISL's pjoject's German grammar
>page [] (and perhaps the Danish one, too),
>you will discover that all indeclinable words are treated as
>particles in terms of word class, or form, regardless whether
>they are adjectives, adverbs, prepositions, nouns, or whatever.
>Not all people would agree on this approach (I certainly wouldn't)...

This suggests a caution to anyone pursuing this issue of indeclinability:
don't take the category assignments in grammars at face value. (It
also raises interesting questions concerning the cross-language
identification of categories/ parts of speech.)

Fred Baube mentioned Finnish as having about "half a dozen" undeclinable
words which are not loans. I have not had the time to look into this.

Nancy Hall wrote about:

>...the "go-see" construction, which only appears
>in cases where a verb can appear as a bare stem. E.g.:
>I want to go see a movie.
>I often go see movies.
>I did / will / should / would etc. go see a movie.
>*I went saw a movie (or go saw, or any other inflection)
>*I had gone seen a movie.
>*He often goes sees movies.

I believe there was a PhD dissertation done on this construction (or else
the "go and see" construction, which Nancy points out does take inflection)
at the U of Washington around 1985. I think the topic has come up once or
twice in Linguist List, and I regret that I don't recall the citation. At
any rate, my suspicion is that this is a different matter from indeclinable
words, but I would like to be proven wrong... "Have got" is another verb in
English which has a defective paradigm (*"Houston, we had got a problem, but
it's fixed now."). There was an article on this in Linguistic Inquiry,
probably in the '70s.

I'm afraid this has been somewhat of a museum of curiosities, with only some
generalizations apparent (at least to me!). I am working (with several
others) on the computational implementation of morphological tools for field
linguists. My query originated in the desire to ensure that we didn't make
it impossible to account for data, even if it has to be done in an ad hoc
way. (Some ad hocness will be necessary, as we boldly go where no theory
has gone before!) The results summarized above did confirm some of my
fears. For example, the fact that some indeclinable Russian nouns can take
derivational morphology, and thereby become inflectable, ruined one solution
I had in mind for indeclinable words. So it's back to the drawing board on
that one. But better to find out during the design phase than after
implementation. Thanks to all!

 Mike Maxwell
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue