Does Jespersen's Cycle apply to languages without negative concord?

Question

In this comment, Rethliopuks mentioned something I'd never really connected in my head before.

[Negative concord] is standard in plenty of languages around the world, incl[uding] most Romance languages, Afrikaans, Greek (Ancient or Modern), nearly all Baltic and Slavic languages, Indo-Iranian languages iirc, Uralic languages, Turkish, etc.

Greek is a language that's famously gone through Jespersen's Cycle more than once, with clear evidence of all the different stages. But it hadn't occurred to me that all those stages had negative concord too. Similarly, the famous example of French ne pas comes from a negative concord language.

Do we ever see Jespersen's Cycle happening in a language without this concord? If so, does the process differ in any significant way? The middle stage of Jespersen's Cycle seems to require multiple negative markers making a net negative, but the Cycle is also talked about as a sort of universal, while negative concord isn't.

---

P.S. I'm not sure if all of my terminology is standard.

To my understanding, in a negative concord language, multiple negatives make each other stronger, and one element of the sentence being marked negative makes the rest be marked negative too, if possible. For example, Yiddish ikh es an epl "I'm eating an apple", but ikh es nit ken epl I eat not no apple "I'm not eating an apple".

In a positive concord language, either you can only have one negative per phrase with everything else being positive (English "I haven't ever done it", *"I haven't never done it"). Alternately, multiple negatives can make a positive: Latin non numquam ēgī not never I.have.done.it = "I've done it at some point".

Answer 1

I think there are two answers that may be helpful here, but before we get to them let's clear up a few things. Negative Concord is not the same thing as Redundant Negation. Negative Concord is a phenomena where two distinct negative items in a language are used for a single instance of sentential negation.

I ain't never been drunk. Interpretation: 'I've never been drunk.' (Alabama English; Feagin 1979)

Redundant negation refers to the use of multiple morphological agreement features for negation as in:

Je ne veux pas manger. Translation: 'I (NEG) do not want to eat' Interpretation: 'I do not want to eat' (Formal Modern French)

If we look at negative concord in French we would see something like:

Personne n'a rien fait . Translation: No one has not done nothing Interpretation: No one did anything (Formal Modern French)

Given that Jespersen's cycle addresses redundant negation not negative concord, the short answer to your question is: they're unrelated issues, redundant negation is distinct from negative concord.

With that in mind we can reframe your question to something similar:

Can Jespersen's Cycle take place without two-part negation at any stage?

There are two possible answers to this question:

1. No, Jespersen's cycle is a process of grammaticalization and accordingly needs both a lexical form that becomes a grammatical one and a lexical form that comes to support it.

   Grammaticalization is a process believed to be constantly at work throughout language whereby lexical forms become grammatical forms. Jespersen's cycle can be usefully seen as the grammaticalization of a negative operator which loses lexical meaning becoming merely a grammatical marker of negation. As such negation requires an additional word which carries the semantic content of a negative operator.

   It follows then that the process described by Jespersen's cycle - like virtually all instances of grammaticalization - necessarily refer to two forms: one item being semantically bleached as it becomes a grammatical marker, that needs the support of another item with the semantic content the original form lost

   Let's look at how this view of Jespersen's cycle would apply to negation in Middle English: Initially 'ne' caries a [+NEG] specification which is interpreted by the semantics as marking negation (Stage I). At some point 'ne' loses this specification and becomes an empty morpheme. Since the language has to have a way of marking negation not is introduced which now carries [+NEG] (Stage II). Even though 'ne' is now semantically empty the morpheme does not disappear until later (Stage III) because loss of morphology lags behind loss of semantics (essentially 'ne' has to die twice once as semantics and once as morphology).

2. Kind-of, there doesn't need to be redundant negation if we don't see what's happening in English as redundant negation.

   Another way of looking at Jespersen's cycle is that it's a single syntactic structure for negation with only part of the structure being overt at each point in time, this assumes that:

   a) Each neg-head must have a neg operator in its specifier b) Each neg operator must be in the specifier position of a NegP that has a neg-head.

   To make this a little clearer let's apply it to Middle English:

   

   we can see 'ne' as the neg-head and ‘not’ as the neg operator. Under this analysis, 'ne' and not are completely distinct and are required to be present in the representation of negation independently of one another. Initially, the element in [spec, NegP] is null and only the head is overt (Stage I). Then the null element in [spec, NegP] becomes overt (i.e. ‘not’) (Stage II). Finally, the neg-head becomes null, leaving only the specifier overt. So, the occurrence of 'ne...not' is simply reflects an overlap between the periods of [Spec, NegP] becoming overt and neg-head not being yet fully changed into a null element across all contexts.

---

So which is it?

As with many things in linguistics it's hard to say for sure. However some interesting work done by Philip Wallage hints at an answer. In the first answer proposed here (grammaticalization) there are two distinct forms of 'ne' at stage I and stage II of Jespersen's cycle respectively. Thus we could predict that these two forms should have different distributions. Looking in corpora Wallage finds (among other things) that the distributions of Stage I and Stage II 'ne' are significantly distinct across middle english. This fact is difficult to account for unless we accept the first explanation presented here using grammaticalization which posits that 'ne' is different at each of those stages.

All of this is to say that it looks like Jespersen's cycle necessarily requires two distinct forms for negation at Stage II, forming redundant negation, however when both are present one form is a grammatical marker and one contains the semantic content for negation.