This has always been a conundrum for me. I like how grammatical case frees up word order, but hate how it puts a limit on what nouns can end in. Years ago someone told me of a natlang that uses 6 declensions, one for each vowel and the 6th for nouns that end in consonants. This allowed nouns to end in any phoneme in the language, though from my own research only plurals were allowed to end in -u. Sadly, I haven't been able to find that language again.

Such a language, if it were to exist, would make it abnormally important to remember each noun's declension class, because what would be the nominative for one noun may be another case entirely for a different noun.

I find it annoying that I can't find that natlang again. I think it was a part of the Slavic family, but I'm not sure. I think it was spoken like somewhere around Eastern Europe or the Caucus mountains.

  • 6
    Are you excluding the trivial options of only having prefixes or infixes?
    – curiousdannii
    Commented Mar 24, 2018 at 9:03
  • Russian?
    – b a
    Commented Mar 24, 2018 at 23:54
  • The natlang sounds like Latin, it has indeed 6 declensions with the described features. However, I'm not sure that it allows any consonant as a final for a noun, at least I can't recall any Latin nouns ending in -b, -d, or -g (stems, yes, but it is like urbs, urbis, laus, laudis or grex, gregis)
    – Sir Cornflakes
    Commented Dec 12, 2022 at 10:50

5 Answers 5


Ignoring the lost natlang, I'll just answer the title question:

There are several possibilities to let nouns end in any phoneme of the language and still have case inflections:

  • Most simple: Have a zero ending in the nominative singular
  • Base your inflection on prefixes (like in Bantu languages) or infixes (like in Semitic languages)

BTW, inflections with thematic vowels (like in Latin and Greek) aren't particularly difficult to remember, because the thematic vowel is easily recognisable throughout the whole inflection paradigm.

PS Many natural languages have only a restricted set of allowed phonemes at the end of a word. Classical Greek is an extreme example of this kind of restriction, but also German has Auslautverhärtung aka final devoicing: Final consonants are always devoiced. This feature does not show up in writing but only in the spoken language.


There are some languages with unusually irregular plurals or verbal forms, but I don't think any language has something quite as stark as what you're demanding here. Mostly because, as you point out yourself, a system that basically makes it impossible to recognize what case a noun is in unless it was previously learned by rote would be so taxing on the mind that it would inevitably regularize to at least some degree within a few generation (French verbs in middle/old French had a very high degree of root alternation, which are now greatly reduced, and English strong verb tend to regularize too).

However have you considered a system of vowel/consonant harmony? I don't think consonant harmony is a widespread feature in any natural language, but for conlanging, the sky's the limit. This would create endings that are affect by the last sound(s) of the root and increase the options.


Here are some possible solutions to this problem that haven't been listed already.

  1. mark case with tone or another suprasegmental feature
  2. case disfixes
  3. have a large class of indeclinable nouns
  4. achieve free word order via noun classes and agreement

1 Case Tone

Maasai marks case with tone. It's predominantly VSO, but frequently uses VOS order if the object is more topical.

2 Case disfixes

Some languages like Alabama (Muskogean, North America) delete phonemes in a root to indicate pluractionality (plural subject/plural object/multiple repetitions), basically verbal number. You can repurpose this feature for case.

Here's a strawman example in a 3-case language (direct, oblique, and genitive).

direct case: unmarked
oblique case: delete the final phoneme
genitive case: delete the final 2 phonemes

Suppose chopal means house.

house                 chopal
to/from/&c the house  chopa
the house's           chop

3 have a large class of indeclinable nouns

You can have a fair number of nouns that don't decline at all before the declension system breaks down completely.

Indeclinable nouns can end in whatever you want them to.

This might work well if your language has optional prepositions or postpositions with the same meaning as certain cases so that the role of an indeclinable noun can be clarified if necessary.

4. noun classes and agreement

Some languages such as Swahili have many noun classes and require verbs to agree with their arguments. Many Bantu languages like Swahili have agreement with direct objects as well, but don't require its usage in all situations.

You can also achieve the same effect with obligatory clitic doubling, such as in Macedonian. The idea is that you must have a pronoun referring to various verb arguments (such as the direct object) even when a full noun phrase is present. Clitic doubling is more effective at disambiguating sentences if you have more noun classes or more grammatical numbers.


Any declensional system that involves suffixes will limit "what nouns can end in", because nouns can only end in the declensional terminations of that language!

Even if you maximise by disallowing repeats, you can come up with say eight stem formations and say eight cases across four numbers. If you make each of these 256 forms "unique" (such that -am never repeats) that's 256 possible syllables.

English has something like ten to maybe fifteen or more thousands of syllables. The number of distinct monosyllabic words in astounding. This source lists hundreds if not several thousand.

I don't see how it could be possible to both create a reasonable declensional system and do anything but limit what nouns can end in.

Any solution to this query will involve something silly like having tens of thousands of discrete morphological forms for nouns. One for each possible syllable. Or, the equally silly notion of restricting the number of valid syllables to, say, 256, thus equalling the number of nominal case endings.

  • It's worth noting though that English (and IE languages in general) is on the higher side of syllable complexity.
    – Circeus
    Commented Mar 25, 2018 at 0:31

In my opinion, the obvious first step is having a base form (be it nominative, accusative, ergative, absolutive or intransitive) with a null ending so that the root form with all its final sound variety actually shows up and can be identified.

Second, instead of attempting to find a declension pattern for each possible final sound, group them together. For example, you might have one (or many) declension groups that include words ending in consonant clusters; their feature may be that the ending tacked on always begins with a vowel to ensure a pronounceable syllable. On the other hand, you might decide to group together words ending in Vs, Vz, Vd and Vt (V = vowel sound) and have all their declension patterns be Vt + ending. These strategies reduce the potentially very high number of patterns to a more managable set – and are likely to have happened in natural languages anyway.

Finally, I would like to point out that Finnish and German are each rather close to what you have in mind. While native Finnish words can only /t, s, n, r, l/ or a vowel, loan words from other languages do not necessarily abide by this rule and newer loans have resisted change to fit Finnish phonotactics somewhat. In declensed forms, an /i/ is added between stem and ending to turn it into a fully allowed syllable. In German (subject to terminal devoicing) there aren’t any restrictions on final sounds that I am aware of. Indeed, declension is usually very simple being made up only of a sometimes optional filler-e plus a case-marking consonant so declension is not as complex as found in other languages.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.