In my attempt to learn Georgian, an agglutinative language of the South Caucasus, I have learned that it has both case endings and postpositions.

I also have some familiarity with Korean and Japanese which are usually analysed as having postpositions (as one subclass of their particle class) but not as having case endings.

When linguists study a new language hitherto lacking any tradition of grammatical analysis, how do they determine whether the language has either or both of these features, and which morphemes belong to which category?

Also when looking at a well studied language are there easy ways to tell the difference between the two or is it at least partly due to the accepted existing analyses?

  • 2
    It looks like the discussion will be similar to an earlier question, "Is there really a difference between agglutinative and non-agglutinative languages when spoken?", since the main issue is whether or not the morpheme is its own grammatical word.
    – user483
    Commented Dec 17, 2011 at 16:03
  • 2
    I see that nobody's mentioned Clitics yet. Cliticization is often a stage in acquiring or losing inflections.
    – jlawler
    Commented Dec 17, 2011 at 17:23
  • 1
    "Today's morphology is yesterday's syntax"
    – Alex B.
    Commented Dec 29, 2011 at 17:45
  • 1
    Case endings are not separate words. (If you can't tell what the words are, that's a different problem.)
    – Greg Lee
    Commented Jul 23, 2016 at 18:15
  • 2
    Following on @GregLee: The definition of "word" is language-specific, but there's usually some sort of phenomena that makes it reasonable to group morphemes into units called words (stress, vowel harmony, pronoun substitution etc.) Then you distinguish inflections from adpositions by whether they're in the same word. E.g. in a vowel-harmonic language, if the morpheme doesn't obey word harmony, it's another word and therefore a postposition. In a language where every word has exactly 1 stressed syllable with a tone, if the morpheme never carries tone, it's an affix. And so on. Commented May 18, 2018 at 12:25

3 Answers 3


Evidently the source you may want for this is the collection "Paradigms: the economy of inflection" edited by Frans Plank, 1991. I don't have access to the full text in electronic form, but in particular the article "Rasmus Rask's Dilemma" provides a list of criteria for distinguishing cases from postpositions. If it's of interest I can request the book from the library and summarize it here. Also see "Noun phrase structure in the languages of Europe" by Plank, 2003.

Also: Friedman in Studia gramatyczne, 1991 has a paper titled "Romani nominal inflection: Cases or postpositions" where he examines this question in Romani (and uses Georgian as one of his models).

Romani is strictly prepositional -- with the exception of five postpositions, which have nearly exactly replaced five original Sanskrit cases. Scholars over the years have variously described these as cases or postpositions. Friedman presents the arguments in favor of and against each interpretation. His conclusion with Romani is that it is best to treat these as case suffixes since (1) they undergo assimilation, (2) the language is otherwise prepositional, (3) the case suffixes have no independent lexical meaning and (4) just because their origin was as postpositions does not matter synchronically. But I recommend reading the whole paper for an excellent example of the reasoning and evidence that went into his conclusions, which is what I believe you are asking for.


This is a fundamental question in morphology that has consequences going far beyond the simple distinction between case endings and postpositions (which, by itself, is effectively quite thorny in many languages). I would say that it pertains to the problem of defining such notions as grammatical category, paradigm and fusional/flective typology in general. Indeed, in order to postulate case endings for a language we should first assure that it has nominal paradigms, which, in turn, implies the existence of inflexional grammatical categories for the noun.

But how would we discriminate between flective languages (i.e. languages with paradigms) and, let's say, agglutinative languages (i.e. those without paradigms)? Some criteria can be suggested here, but the list can be easily emended or enlarged. For the sake of simplicity I am exemplifying every statement with facts from nominal inflexion, even if this approach is valid for the paradigmatic morphology in general.

  1. Grammaticality implies obligatoriness: cases are obligatory on all nouns in all instances. Such grammatical categories as case or number or gender can't be freely avoided by the speakers. In agglutinative languages this is not always the case: sometimes no grammatical marker is also an admissible option (this happens with Japanese postpositions, especially in the spoken language).
  2. Grammatical categories imply the possibility of making a grammatical error, since they are selected syntactically: if you say many mouse it would be clearly an error that any speaker of English will be able to emend into the correct form many mice, which fits better the syntactic context. The postpositions, on the other hand, are not selected strictly syntactically since they can vary according to the communicative intentions of the speaker. Therefore, there is no such thing as "wrong postposition" (if not stylistically).
  3. Grammatical categories imply morphological variability and inflectional classes. Indeed, in a language that don't vary its case endings, you basically can't tell the difference between cases and postpositions, since both will appear as grammatical morphemes without any variation. On the other hand, in prototypical flective languages, cases are instantiated by sets of allomorphs, which are either contextual (i.e. depend on the phonological form of the stem) or, importantly, non contextual (i.e. are lexically bound, forming inflectional classes, such as the five nominal declensions in Latin).
  4. Grammatical categories are better when they are semantically uniform, i.e. refer to a set of uniform meanings. Thus, cases can apparently convey many different meanings, but such meanings would not usually overlap with other semantic content, like gender or number or, say, tense (i.e. in languages with cases, no "feminine", "future" or "plural" case would be, normally, attested). On the other hand, postpositions can bear many different functions, from syntactic relations to honorific connotations.
  5. Cases usually belong to cumulative morphemes which also include gender, number and possibly some other grammatical categories; the postpositions, instead, are simple morphemes by definition.
  6. Some minor features can be added:
    • regularity of markedness (e.g. case do not usually allow markedness shifts: if the nominative singular case is unmarked in some nouns, then the dative plural or the instrumental dual cannot be unmarked in other nouns; however, this rule is less strict; contrariwise, let us think to the Japanese verbal transitivity markers: some verbs have unmarked intransitive form, while some others have unmarked transitive form, which strongly suggest that transitivity is a derivational category, rather than an inflectional one;
    • you usually cannot have multiple case markers at the same time; on the other hand, you can have serialisation of postpositions (think of such Japanese NP structures as N ni wa, N de no, N no ga, and similar);
    • case categorial values usually don't exceed a dozen; systems with up to 40 cases are obviously agglutinative.

The labels "flective" and "agglutinative" may be as antiquate and outdated as you want; however, defining a prototypical flective language (vs. a prototypical agglutinative language), on the ground of the above mentioned criteria, provides a useful analytical tool. Thus, a prototypical flective language will have: rigid paradigms defined by the cumulative marking of two or more obligatory, error-provoking, syntactically selected grammatical values, with non contextual allomorphy, semantic uniformity, no shifts of markedness, no serialisation, no optionality, and with a limited number of grammatical values.


for example see what i have written about tatar language: https://sourceforge.net/p/apertium/mailman/message/32515613/ .

it is even hardedr in tatar language, because it is not inflectional, but agglutinative language, but even in it, they can be divided into "cases" and "pospositions".

but as i remember i tend to think that they should not be divided in tirkic languages, though i agree that there are several types of them , some are like case, some are like preposition.

i think, most easy distinction in tatar is that there maybe several "cases" which behave like cases with prepositions, for example, genitive case is used with many prepositions.

in fusional inflection languages "cases" can be also divided because of being fusional.

i have checked 11 parameters in tatar language, you can see a table.

and tatar language has traditional grammar, and i very dislike it.

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.