Does anyone know about the development of agglutinative languages out of fusional languages, or, more precisely, agglutinative features out of fusional features? I am thinking in particular about the development of Nepali out of Sanskrit or the Sanskrit-derived fusional language that was a mother to Nepali.

In Sanskrit and other older fusional Indic languages, you have case endings which vary depending on the stem; in Nepali there is a regular set of post-positions, regardless of the stem word:

-le – "by" (instrumental)
-lāī – "for" (dative)
-bāṭa – "from" (ablative)
-ko – "of" (genitive)
- – "in," "at," etc. (locative)

Whereas in Sanskrit, and presumably in the Middle-Indic language that directly produced Nepali, the case endings will vary depending on the stem, for example:

Skt.: hastāt "from the hand" ( āt ablative after -a stem hasta)
Nep.: hātbāṭa

Skt.: hastinaḥ "from the elephant" (-(n)aḥ ablative after -i(n) stem hasti(n))
Nep.: hastibāṭa

Skt.: devyāḥ "from the goddess" (-āḥ ablative after -ī stem devī)
Nep.: devībāṭa

It seems this may be a common occurrence, a fusional language naturally developing into an agglutinative language. Or is there influence from other (quasi-)agglutinative languages (e.g. Newari or Tibetan) that may explain this?

  • I thought in the cycle hypothesis fusional languages developed from agglutinating languages, not the other way round.
    – curiousdannii
    Sep 4, 2020 at 22:55
  • Hmmm. I'm no expert myself, but in the case of Nepali (and Hindi), there is no doubt that they are more agglutinative than their parent Middle Indic languages, which are more fusional.
    – Paul
    Sep 4, 2020 at 23:54
  • @curiousdannii: The typological cycle theories are just now being deconstructed. There are transition probabilities between any possible typological status, and the transitions aren't really circular. Sep 18, 2020 at 22:45

2 Answers 2


The terminology gets in the way a bit, because agglutinative, analytic, fusional etc. are not really crisp and well-defined states for languages to be in, rather, the re is a continuum from 1-to-1 form/function correspondence to many-to-many relations, and a continuum of using morphological vs. syntactic means of expressing things. There is a tendency for certain kinds of words in sequence to become affixes, often termed "grammaticalization". This especially involves words that take on special grammatical function, like "of" or "not" in English, or "ni" (appr. "is") in Bantu. Things that are called "clitics" such as contracted "n't" in English or apostrophe-s for possessive typically reflect an intermediate stage, where the word acts more like a part of the word that it attaches to.

Fusional systems like in Sanskrit tend to simplify, by eliminating the difficult-to-parse combinations e.g. person, number, tense, mood in Sanskrit verb inflections, so that the inflectional system of Middle Indo-Aryan becomes simpler in the development from Ardhamagadhi to Apabhraṃśa, then the modern languages. I think the best way to treat that progression is in terms of the parsing problem of figuring out many-to-one meaning-to-form correlations in Earlier Indic (or other languages with similar properties), and the natural tendency to conventionalize meanings of multi-word expressions (like "got taken", "have seen") and glue them onto the head word as a new kind of inflection.

To take an example from the Bantu language Shona, a number of prefixes have been historically added which qualify the action performed, and which derive from merging auxiliary-type verbs with a main verb, for example -wiriro- 'happen to' ← 'fall', -zo- 'remote' ← 'come', -do- 'polite' ← 'want'. These can be piled up, e.g. ku-zo-wiriro-do-bhadhara "to maybe happen to pay (please)".

  • 1
    Great answer, thanks! I'm struggling to understand the meaning of this sentence: "I think the best way to treat that progression is in terms of the parsing problem of figuring out many-to-one meaning-to-form correlations in Earlier Indic (or other languages with similar properties." Do you mean by this, the tendency to move from "many meanings–one form" towards "one form–one meaning," which in the case of Skt./Nep. would be the meanings denoted by the forms of noun and verb endings/post-positions?
    – Paul
    Sep 4, 2020 at 21:09
  • For example, Skt. case suffixes are kind of incoherent in having multiple cases represented by a single suffix, but only in the dual, or only in the plural. One solution is to get rid of the dual, another is to get rid of the elaborate case system in favor of a smaller "nominative/oblique" system. Though to be more accurate, cases are primarily syntactic properties and not just about semantics, so I should say "function" rather than "meaning".
    – user6726
    Sep 4, 2020 at 21:14
  • I see. Now I will try to determine if there are stock phrases that may have become fossilized as the various post-positions in Nepali. One possibility I see for the Nepali dative "lāī" is Skt. "arthāya," although "rth" > "l" seems implausible.
    – Paul
    Sep 4, 2020 at 21:25

Looking through Turner's etymological dict. of Nepali, I have some more ideas:

(dative) lāi < Skt. lābhe; or infinitive of Skt. lāgayati (Pk. lāiuṁ, Ap. lāivi; v.s.v. lāunu). See J. Bloch, p. 202.

(ablative) bāṭa < bāṭo "way" < Skt. vartman

(genitive) ko < Skt. kṛtya

(locative) mā < Skt. madhya

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.