Factores psicológicos

Languages vary with respect to whether they allow multiple apparently inherent- ly negative constituents to appear, say, within the same clause, without canceling each other out. Van der W ouden (1994: 95) distinguishes between two phenomena that he labels “negative spread” and “negative doubling”. Negative17 doubling is illustrated in (14a) and (15a) in the NC languages Spanish and Ital- ian respectively. Here, a negative XP, 18 nadie/nessuno ‘no-one’, appears

(iv) A nessuno Gianni (’??non) telefona. (Haegeman 1995: 196 (43c)) to no-one G. non telephones

‘G. doesn’t call anyone.’

In similar configurations in Spanish, negative doubling is not attested:

(v) A ninguna de ellos (’no) llamaría yo. (Suñer 1993: 3) to none of them no call-COND I

‘I wouldn’t call any of them.’

For cross-Romance discussion of NC, see Posner (1996: 148!49, 302!5).

19. Etymologically, Spanish n-words are not negative (Laka 1993a). For contemporary speakers, however, one could plausibly assume that these elements are analyzed as being inherently (morphologi- cally) negative. The situation in Italian is different; for Acquaviva (1995: 14fn3), nessuno, for example, is derived from the Vulgar Latin ne-ipsu-unu ‘not-even-one’, in other words etymologically negative.

together with)“doubled” by)the regular pre-verbal negative marker, namely no/

non. Examples (14b) and (15b) illustrate negative spread, whereby multiple negative XPs co-occur. Examples (14c) and (15c) show negative spread and negative doubling occurring simultaneously.

(14) a. No conozco a nadie. (Spanish)

no know-1SG no-one ‘I don’t know anyone.’ b. Nadie me ha dado nada.

no-one me has given nothing ‘No-one has given me anything.’ c. No doy nada a nadie.

no give-1SG nothing to no-one ‘I’m not giving anything to anyone.’

(15) a. M ario non ha visto nessuno. (Italian) M . non has seen no-one

‘M . hasn’t seen anyone.’ b. Nessuno ha fatto niente.

no-one has done nothing ‘No-one did anything.’

c. Gianni non dice niente a nessuno. G. non says nothing to no-one ‘G. doesn’t say anything to anyone.’

T he crucial property of all the examples in (14) and (15) is that, although all the italicized constituents (the regular negative marker and the negative X Ps) are ar- guably morphologically negative, each sentence is interpreted as a single19 instance of sentential negation. They are not interpreted as containing multiple instances of logical negation. Spanish and Italian are NC languages.

Languages that do not allow multiple occurrences of negative constituents to be interpreted as a single instance of sentential negation are termed non-NC

languages. In these languages, where two negatives co-occur, the first negation takes scope over, and cancels, the second. Examples are SE and German:

(16) a. I’ve (’not) seen no-one. (SE)

b. I’ve % (’not) given nothing to no-one.

(17) a. Hans sieht niemanden (’nicht). (German) H. sees no-one not

‘H. can’t see anyone.’

b. Ich bin mit niemandemnirgendwohin % (’nicht) gefahren. I am with no-one nowhere not travelled ‘I didn’t drive anywhere with anyone.’

The SE examples in (16) show that one or more negative XPs cannot co-occur with the verbal marker of negation not (and receive the relevant NC interpretation). SE does not, therefore, demonstrate negative doubling. W ith respect to negative spread, the issue of whether m ultiple negative XPs can co- occur is unclear. Some speakers as well as prescriptivists reject (16b) without

not; others do not. Dialectal variation seems to be at play. Clearly, though, negative spread needs to be distinguished from negative doubling. The judgments in (17) suggest that German patterns essentially with SE in this respect.

Ascertaining whether the Negative Cycle is relevant to NC amounts to establishing whether there is a correlation between (a) whether a language is an NC language and (b) where it stands in the Negative Cycle. Such a correlation would suggest that one was determined by the other. Given considerations of learnability, this would presumably mean that where a language stands in the Negative Cycle determines whether or not it is an NC language. Jespersen (1924: 333) suggests that there is such a correlation:

There is one very important observation to be made, without which I do not think that we shall be able to understand the matter, namely that repeated negation [i.e., NC] becomes an habitual phenomenon in those languages only in which the ordinary negative element is comparatively small in phonetic bulk. . . . If this repetition is rarer in modern English and German than it was formerly, one of the reasons probably is that the fuller negative

not and nicht have taken the place of the smaller ne and en.

Following my analysis of the Negative Cycle within the framework of the NegP hypothesis in section 3.1.2, I shall assume that Jespersen’s observation amounts

20. Note the following observation made by Acquaviva (1993: 60!61): “We can now formally characterize the difference between the English and the Romance (and nonstandard English) operators: only [in] the latter are specifiers of heads endowed with the morphological negative feature.”

My analysis of Jespersen’s Generalization differs in crucial ways from Acquaviva’s; this is discussed in section 3.4.

21. Note that my interpretation of Jespersen’s observation is slightly different from the one tentatively proposed by Haegeman (1995: 165). In the context of Jespersen’s observation, Haegeman suggests that NC may be determined by the availability of an overt negative head. See also Haegeman (1991: 16):

We might propose that in languages with NC readings the head of NegP is “strong”: it is auto- nomously licensed: it has its NEG feature in the base. The Neg Criterion is met by a “strong” static agreement configuration. In non-NC languages, on the other hand, Neg is “weak” and would be assigned the NEG feature by its specifier by virtue of spec-head agreement. . . . What is crucial for NC . . . is that the NEG feature on NegE is independently licensed, i.e., that NegE

is a strong head. In languages where the NEG feature on NegE can only be achieved via dyna- mic agreement the negative head is not strong and NC is not possible.

For me, in contrast, NC correlates with the absence of a negative operator in SpecNegP. Assuming that the characterization of the data in sections 3.1 and 3.1.2 is correct and that languages can indeed mark sentential negation by overt material associated with both SpecNegP and NegE, the difference between Haegeman’s and my own (re-)interpretation of Jespersen’s observation is not a trivial one. Note also that Haegeman’s approach predicts that Modern French is a non-NC language. See section 3.5.2.

22. The consequences of the spec-head agreement mechanism inherent in the Neg Criterion not- withstanding, which is, in any case, discussed in section 3.3.2.

to what I referred to as Jespersen’s G eneralization in (2), repeated here for20 convenience:

(18) Jespersen’s Generalization:

A language is an NC language iff the regular marker of pure sentential negation is not associated with SpecNegP.21

3.2.1