U N I V E R S I T A T E A D I N B U C U R E Ş T I F A C U L T A T E A D E L I M B I Ş I L I T E R A T U R I S T R Ă I N E

PROIECT DE LICENŢĂ

CANDIDAT: Sandra – Iulia RONAI

ÎNDRUMĂTOR ŞTIINŢIFIC: prof. dr. Larisa AVRAM

– 2010 –

B U C U R E Ş T I

U N I V E R S I T A T E A D I N B U C U R E Ş T I F A C U L T A T E A D E L I M B I Ş I L I T E R A T U R I S T R Ă I N E

N E G A T I O N AND RIZZI’S LEFT PERIPHERY

– EVIDENCE FROM ENGLISH AND LATIN –

Negaţia şi periferia stângă a lui Rizzi

– Argumente din engleză şi latină –

CANDIDAT: Sandra – Iulia RONAI

ÎNDRUMĂTOR ŞTIINŢIFIC: prof. dr. Larisa AVRAM

– 2010 –

B U C U R E Ş T I

TABLE OF CONTENTS

Introduction 1

1. Rizzi’s Split-CP Hypothesis 2

1.1. Introductory remarks 2

1.2. The split CP – Rizzi’s Left Periphery 4

1.3. The Topic – Focus system 6

1.4. The Force-Finiteness system 9

2. Negation and Double Negation 11

2.1. Types of Negation 11

2.1.1. The Scope of Negation 11

2.1.2. Tests for sentence negation 13

2.2. Polarity and Polarity Items 14

2.3. Negative elements 16

2.4. Interpreting multiple negative elements 17

2.5. A Short History of Double Negation 18

2.6. Conclusions 26

3. An overview of negation in Latin 27

3.1. Negative Elements in Latin 28

3.1.1. Negative markers 28

3.1.2. Semi-negatives 32

3.1.3. N-words 33

3.1.4. Negative quantifiers 33

3.2. Negative Polarity Items 35

3.3. Double Negation 38

3.4. Conclusions 42

4. Arguments in favor of Negation at the Left Periphery 42

4.1. Scopal Ambiguities 43

4.2. Negation as the Binding of Free Variables 45

4.3. Semantic Negation in Non-Negative Concord Languages 46

4.3.1. Syntactic vs. Semantic Negation 46

4.3.2. Negation as Feature-Checking 47

4.4. Negation at the Left Periphery 49

4.4.1. Theoretical arguments 50

4.4.2. Empirical evidence: Latin and other languages 52

4.4.3. Extending the analysis: English 57

4.5. Conclusions 59

Final Conclusions 59

Bibliography 60

1

Introduction

This paper grew out of my desire to blend together my two main academic

interests: on the one hand, an old and enduring passion for the many intricacies of the

Latin language; on the other, a new and growing enthusiasm for Generative

Linguistics. Ever since my first contact with the concepts of Generative Grammar, I

dreamed to apply the theoretical framework that I was learning to the systematic

analysis of empirical evidence found in Latin – especially since, unfortunately,

modern linguistics pays too little attention to the Classical languages.

When I started studying negation in English, I was immediately charmed: the

topic promised to raise exciting challenges. For one thing, I was intrigued by the

unexpected similarity between the mechanism for expressing negation in English and

what I knew about the type of negation in Latin. How come the two unrelated

languages lack Negative Concord, whereas Romanian and the Romance languages are

Concord languages? And what is, essentially, the distinction between Negative

Concord and Double Negation?

These questions, and many more, prompted me to start exploring negation,

back in the spring semester of my second year. The first thing that I could clarify was

the surprising evolution of Double Negation in English and Latin. This initial research

resulted in a short paper presented at our Faculty’s Students’ Conference, in 2009;

some useful material found its way into this paper as well, and was integrated in

chapter 2.5. But the most important thing that I gained from that short paper was a

desire to delve deeper and discover the answers that I was not able to find at that time.

I just knew that my diploma paper would deal with negation in English and Latin.

Negation, however, is a very vast subject matter, and soon I realized I had to

focus on something smaller – like, for instance, the issue of negative

complementizers. I remembered that Latin could convey sentential negation by means

of a negative particle that replaced the affirmative complementizer or even the

affirmative copulative conjunction. What could this particle be? How could it express

negation? And most of all, where could it stay in the syntactic tree? This paper is an

attempt to answer these questions.

In order to solve this mystery, I had to first assume a theoretical framework

that accounts for the complementizer layer in detail – therefore I read about, was

convinced by, and adopted Rizzi’s Split-CP hypothesis. This is the subject of

2

Chapter 1 of my paper. The next thing I needed was to refresh my memory with

respect to different major topics in the study of negation in general. I present some of

them in Chapter 2. Then, a detailed description of negation in Latin was in order – I

deal with this in Chapter 3. At last, in Chapter 4, I bring a few more supporting

arguments, then, finally bringing everything together, I suggest an analysis of

negation in the Left Periphery, for the two languages under scrutiny.

1. Rizzi’s Split-CP Hypothesis

1.1. Introductory remarks

In order to establish the exact place of the NegP in the syntactic trees for

English and Latin, we first have to take a closer look at the general structure of the

clause, as envisaged in the Generative framework.

The central unit of this structure is the well-known X-bar schema, which I

repeat here for convenience:

(1) XP 3Spec X’YP 3

X0 Compl ZP

As we can see, the structure of the phrase includes a maximal projection XP,

the intermediary node X’ and the head X0. Other maximal projections, themselves

with a similar structure, can appear in the Specifier and in the Complement positions.

It is essential to note that this schema is applied to both lexical and functional

projections, an observation that will prove to be of crucial importance for our analysis.

Actually, as Rizzi (2000) clearly states, it was noticed that the X-bar schema is

sufficient to accommodate all the information that makes up the structure of the

clause:

3

“...all structural configurations are headed, and the only way to produce

structure consists in starting from a head (lexical or functional) which projects

a phrase in accordance with the fundamental X-bar schema and which can

recursively take other phrases as specifiers and complements.” (Rizzi 2000:7)

Therefore, the basic structure of the clause in X-bar Theory is this:

(2) CP 3

C’3

C0 IP 3

I’3I0 VP

3 V’

3 V0

Here, each one of the three phrases represents one of the three layers of the

clause. The most important of the lexical projections is the VP. Above it, we have

Inflection, an abstract, umbrella-like projection for all the “concrete or abstract

morphological specifications on the verb” (Rizzi 2000: 260). Finally, above IP, there

is the CP, where we can usually find what traditional grammars call “link words” –

complementizers such as “that” and “for” or relative pronouns and adverbs (also

called Wh-words). Because, as Rizzi (2000: 260) explains, each clause is assumed to

consist of three such structural levels:

1. a lexical layer, with V0 as its head,

2. an inflectional layer (the IP projection) and

3. a complementizer layer.

Concerning this last layer, Rizzi notices that, apart from the operator-like

elements that I have already mentioned above, the Complementizer also includes

topics and certain focalized elements. In fact, he proposes a “fine-grained” structure

of the CP layer, which I will be adopting in this paper.

4

1.2. The split CP – Rizzi’s Left Periphery

Even if the representation in (2) is correct and very clear, in time, it was felt to

be “too simplistic” and insufficient for a precise description of the structure of a

clause. The first of the layers to be re-analyzed in terms of a number of smaller

functional projections was the IP layer. The too general and too vague name of

“Inflection” was later replaced by separate projections for each of the morpho-

syntactic features of the verb (like tense, aspect, mood, agreement, etc), so the

inflectional layer was designed as consisting of TP, AspP, MoodP, AgrP and so on,

depending on the language in question and on how exact the analysis wanted to be.1

Two names that Rizzi mentions as having brought a great contribution in

accounting for the detailed description of the inflectional space are those of Jean-Yves

Pollock and Gugliermo Cinque. In his 1989 essay on verb movement, Pollock proved

that, by postulating different distinct inflectional heads, it was possible to solve two

important issues: first, it could explain various correlations between form and position

in the verbal system, “thus grounding the study of verbal morphology on principles of

syntactic organization” (Rizzi 2000: 7) and second, it made it possible to justify

various adverbial positions. Cinque (1999), in turn, came up with an intricate

cartographic description of the functional structure of the clause. Because of the

positive results obtained by splitting the IP, “It was only natural”, says Rizzi, “to

extend this line of inquiry to other categories”.

Following the same guidelines that led to establishing the fine structure of the

inflectional system, Rizzi proposes an equivalent analysis for the complementizer

layer. The theory that I will summarize here was articulated by Luigi Rizzi in his

paper The fine structure of the Left Periphery, written in 1995 and first published in

19972. For Rizzi, the Left Periphery includes all the elements that appear before the

IP, therefore to the left, in the syntactic configuration.

The criteria that he took into consideration when proposing his system were

twofold: on the one hand, there are word order considerations, which directly lead to

detecting positions, since they license heads which serve as landing sites for elements

1 In fact, for English, conceiving separate projections within the IP seems somehow redundant, due toits much reduced morphology – therefore the representation of English clauses can still feature theunique IP. However, for languages with a rich morphology, like Romanian (or, of course, Latin), thesplit IP hypothesis has to be adopted.2 I had access to the version reprinted as Chapter 10 of his book Comparative Syntax and LanguageAcquisition.

5

moved to the Left Periphery – because “all kinds of movements to the left periphery

must be motivated by the satisfaction of some criterion, hence by the presence of a

head entering into the required Spec-head configuration with the preposed phrase”

(Rizzi 2000: 261); on the other hand, there are some indirect “subtler interactions with

principles of structural organization and well-formedness” (Rizzi 2000: 8).

On the basis of complex studies, first on Italian and then extended to other

languages, Rizzi identifies two systems which make up the CP space:

1. the Force-Finiteness system and

2. the Topic-Focus system,

each with at least two separate positions. These positions consist of different

functional heads, which may either be phonetically realized or remain phonetically

void, but are still present in the structure. The Specifier position of these phrases has

its role as well, as it is here that some elements are placed – for instance, focalized

constituents have to move up to the Spec of Focus, while the head of such a maximal

projection is filled by the silent operator Focus.

The order in which these phrases appear is fixed and constitutes the precise

structure of the complex CP. For Italian, and, with limited parametrizations, for all the

other languages as well, the cartography of the complementizer layer is eventually

articulated as follows:

(3) Force Topic* Focus Topic* Finiteness IP

It is beyond the scope of my present paper to strictly pursue Rizzi’s line of

thought and present again all the evidence that he puts forth in favor of his hypothesis,

especially the various adjacency and anti-adjacency effects that appear at the

interaction between the C system and the different types of subjects (overt DP, trace

or PRO). Rizzi takes all this into account (2000: 280-307) and uses his observations as

arguments supporting his proposal. For the purpose of this paper, I will be taking his

description of the Complementizer layer for granted and only discuss it as far as it

concerns our analysis of the possible locus of Negation.

6

1.3. The Topic – Focus system

Unlike Rizzi, I will be starting with the part of the Complementizer dealing

with topicalized and focalized elements, “which are by and large independent from

selectional constrains” (Rizzi 2000: 264). Both Topic and Focus3 are conceived of as

separate instantiations of the X-bar scheme, having the following representations:

(4) TopP (5) FocP3 3

XP Top’ ZP Foc’ 3 3Top0 YP Foc0 WP

XP = topic ZP = focus

YP = comment WP = presupposition

To explain the analysis, here are two examples that Rizzi gives:

(6) Your book, you should give t to Paul (not to Bill). (= Topic)

(7) YOUR BOOK you should give t to Paul (not mine). (= Focus)

Even if the two English sentences are formally very similar, their

interpretation (aided by some para-linguistic factors such as intonation, pauses and

stress) is different. In the case of (7), we are dealing with a piece of new information

that the speakers brings into the conversation, whereas in (6) it is “old information,

somehow available and salient in previous discourse”, with the comment (i.e., what

comes after the coma in the sentence) being “a kind of complex predicate, an open

sentence predicated of the topic and introducing new information”. On the contrary, in

(7), “the open sentence expresses contextually given information, knowledge that the

speaker presupposes to be shared with the hearer” (Rizzi 2000: 264). Despite the

similarity the two types of constructions seem to display in English, at a closer

3 Just like Rizzi, I will only be dealing with the pre-posed (usually contrastive type of) Focusconstituents, leaving aside the situations in which a focalized element remains in situ.

7

analysis they prove to be, in fact, the opposite of one another, from a discourse-

analysis point of view. I tried to organize these properties in the following table:

Table 1: Properties of Topic and Focus (1)

is placed…

is marked in

pronunciation by…

contains

information that is

is followed by a clause

containing information that is

TOPIC

… at the Left

Periphery

… a “coma-like”

pause … old … new

FOCUS

… at the Left

Periphery … stress … new … old

In other languages, such as Italian, the difference between the Topic structure

and the Focus one is visible in their syntactic structure as well: in the examples below

(Rizzi 2000: 265), the clitic pronoun lo appears only in the topicalized construction,

but is unnecessary (and incorrect) in the Focus situation:

(8) Il tuo libro, lo ho letto. (= Topic)

‘Your book, I have read it.’

(9) IL TUO LIBRO ho letto (, non il suo). (= Focus)

‘Your book I read (, not his).’

The absence / presence of this “resumptive clitic” (and the reasons behind it)

are just one of the many arguments that Rizzi employs in order to provide an

empirical basis for his hypothesis. Another very straightforward and easy to

understand piece of evidence comes from the use of bare quantificational elements –

these can be present in Focus constructions but are disallowed in Topic (Rizzi 2000:

270):

(10) a. * Nessuno, lo ho visto. (= Topic)

‘No-one, I saw him.’

b. * Tutto, lo ho fatto. (= Topic)

‘Everything, I did it’

8

(11) a. NESSUNO ho visto t. (= Focus)

‘NO-ONE I saw.’

b. TUTTO ho fatto t. (= Focus)

‘Everything I did.’

A third argument (and the final one that I will be discussing here) concerns the

property of uniqueness. While a clause can include “as many topics as are consistent

with its (topicalizable) arguments and adjuncts” (Rizzi 2000: 270), the Focus

projection has to be unique. In syntactic terms, we say that TopP is recursive, but

FocP is not (Rizzi 2000: 276). The simple, unscientific explanation would be that we

can place as many constituents at the start of an utterance and still understand its

meaning, but we can draw the attention of the interlocutor to only one of them (or, in

other words, we have to focus on one single element).

Rizzi, however, also has another, more systematic account. It starts from the

already-observed fact that the Complement position of the Focus projection (WP in

(5) above) has to be occupied by old, given information. If we are to imagine an

utterance with a repeated Focus phrase, the structure would involve the second FocP

instead of the WP. Since Focus implies new information, and ComplFoc presupposes

old information, we would have an “interpretive clash” (the same information would

have to be, at the same time, both old and new) that bans the recursion of the Focus

projection (Rizzi 2000: 277). Such problems do not arise in the case of Topics:

“nothing excludes that a comment (the complement of the topic head) may be

articulated in turn as a topic-comment structure, so that the topic phrases can undergo

free recursion” (Rizzi 2000: 277).

We now have three more differences between Topic and Focus, summarized in

the following table:

Table 2: Properties of Topic and Focus (2)

followed by resumptive

clitics…

bare quantificational

elements… uniqueness / recursion:

TOPIC YES … cannot appear … is recursive

FOCUS NO … can appear … is unique

9

1.4. The Force-Finiteness system

I consider these specifications to be sufficient for the purpose of my paper.

Instead of following the rest of Rizzi’s arguments, I will now turn to the other sub-

system of the Complementizer layer, that which includes specifications for Force and

Finiteness.

Before that, it is time to address a trivial question which I have been avoiding

so far: what is, actually, the role or purpose of this Complementizer layer in the

structure of the clause? Rizzi (2000: 262) has the answer:

“We can think of the complementizer system as the interface between a

propositional content (expressed by the IP) and the superordinate structure (a

higher clause or, possibly, the articulation of discourse, if we consider a root

clause). As such, we expect the C system to express at least two kinds of

information, one facing the outside and the other facing the inside.”

This “information looking at the higher structure” (Rizzi 2000: 262) is related

to the “clausal type” (declarative, interrogative, exclamative, relative, comparative

etc.) or, in Chomsky’s (1995) own terms, to the “specification of Force”. Force may

or may not be morphologically encoded in an overt way: there can be special

morphology for declaratives or interrogatives, or there may be just the required

structure to host a specific operator, or, rarely, both.

As for the inside-facing component of the CP, this is related to the content of

the IP that it precedes, in accordance to some well-documented rules of ‘agreement’

between C and I – for instance, English “that” selects a tensed verb, while “for” co-

occurs with an infinitive; in Romanian, “să” introduces a Subjunctive mood, and “că”

always precedes an Indicative mood. These specifications are found to be all related

to the inflectional category of Finiteness (Rizzi 2000: 263).

The Force – Finiteness system is the obligatory component of the

Complementizer layer: Finiteness links the CP to the IP embedded under it, and Force

‘sets the tone’ of the utterance, connecting it to other clauses or, in the case of root

clauses, to the larger structure of the discourse. Therefore, the Force – Finiteness

system is always needed and always present.

10

“On the other hand (argues Rizzi) it is reasonable to assume that the

topic-focus system is present in a structure only if ‘needed’, i.e., when a

constituent bears topic or focus features to be sanctioned by a Spec-head

criterion. If the topic-focus field is activated, it will inevitably be ‘sandwiched’

in between force and finiteness, as these two specifications must terminate the

C system upward and downward, in order to meet the different selectional

requirements and properly insert the C system in the structure.” (2000: 267)

Taking all this into consideration, we can say that we have justified the “fine

structure of the Left Periphery” that I adopted here. The proposed structure, suggested

as (3) above, and repeated in (12) for convenience:

(12) Force Topic* Focus Topic* Finiteness IP

has now been explained. Therefore, we can at last illustrate the final representation of

the Complementizer layer, as envisioned by Rizzi (2000: 277):

(13) ForceP 3

Force’ 3Force0 TopP*

3Top’

3 Top0 FocP

3 Foc’ 3

Foc0 TopP*3

Top’ 3 Top0 FinP

3 Fin’

3 Fin0 IP

11

2. Negation and Double Negation

In this section of my thesis, I will try to outline a few of the most important

aspects in the linguistic study of negation, providing some working definitions for

terms and categories to be used in the remaining of this paper, and also offering some

necessary clarifications and explanations of the phenomena described. The object of

study is negation seen from a strictly linguistic, and mainly syntactic, point of view –

considerations related to the various problems that negation raises in the domains of

logics, philosophy or psychology are not within the scope of my research; pragmatic

issues will be kept to a minimum as well.

The chapter is organized as follows: in 2.1, I present a classification of

negation with respect to its scope, reviewing some of the syntactic tests used for

differentiating sentential negation in 2.1.2. Sub-chapter 2.2 is a brief presentation of

the concepts of polarity and polarity items, and 2.3 includes a definition and a

classification of negative elements. In 2.4, I adopt a classification of languages

according to the type of negation they display, while section 2.5 is dedicated to

suggesting an evolution of Double Negation in English and Latin. I conclude with a

summary, in 2.6.

2.1. Types of Negation

Negation can be, and has been, classified according to different criteria.

Huddleston and Pullum (2002: 787-790) adopt a framework involving four major

contrasts, separating verbal from non-verbal negation, classifying verbal negation in

English as either analytic or synthetic, explaining the difference between clausal and

sub-clausal negation and, finally, talking about ordinary vs. metalinguistic negation.

I will try to confine myself to a purely syntactic point of view.

2.1.1. The Scope of Negation

Whenever a negative element appears in a sentence, there are several

possibilities with respect to the range of its “influence”. The negator can “affect”

12

(henceforth, ‘take scope over’) either the whole clause, or just a part of it (local or

sub-clausal negation). We distinguish4 between three types of negation:

1. word negation (also called affixal negation) – negation takes scope

over a single word and is expressed by means of an inherently negative affix,

generally a prefix: ‘unwanted’, ‘to displease’, ‘inefficient’, Rom. nefericit,

ilegal, amoral, etc.

2. phrasal or constituent negation – it takes scope over a certain phrase

but does not render the whole sentence negative:

(1) a. He lives [not very far away] from his fiancée.

b. M-am întâlnit cu el [nu foarte demult].

‘I met him [not very long ago].’

The fact that it is the whole phrase that is under the scope of the negation

can be proved by moving the constituent to a topicalized position:

(2) a. She would feel very alone [with no friends nearby].

b. [With no friends nearby], she would feel very alone.5

3. sentential (or clausal) negation – the negative element takes scope

over the entire clause. This used to be called “nexal negation” (Jespersen

1917), understanding it as a negation of the link (“nexus”) between the subject

and the predicate. According to Cornilescu (2003: 38), “a sentence is negative

when its predicate is negated, in other words, when its Inflection, which is the

head of the sentence, is negative.”

It is exactly this type of negation that will be our main concern from this point

on. But since sentential negation could be confused with instances of sub-clausal

4 As in Avram (2009) or Cornilescu (2003: 37-38)5 The difference between this sentence, whose sub-clausal negation does not trigger subject-auxiliaryinversion, and a possible (2. c) sentence, With no friends nearby would she feel alone, is discussed byHaegeman (2000), who argues that (2. b) is a case of topicalization, while the sentential negation in (2.c), which needs SAI, is placed in a Focus position.

13

(affixal or phrasal) negation, we need a way to make sure that negation takes scope

over the entire clause. This can be ascertained by using a battery of syntactic tests,

first proposed, for English, by Klima (1964).

2.1.2. Tests for sentence negation

In his paper, Klima introduced the testing mechanism that has been used ever

since for determining the scope of negation. The following tests are valid for English,

but, for other languages, linguists have to suggest equivalent syntactic devices. We

can group the tests as follows:

1. tag question tests

2. the either / neither tests

3. the not-even test.

1. In English, the polarity of the tag question has to be the reverse of the

polarity of the main sentence: negative sentences are followed by an affirmative tag,

while affirmative sentences take a negative tag:

(3) a. He dislikes coffee, doesn’t he? (affirmative sub-clausal negation)

b. *He dislikes coffee, does he?

c. He doesn’t like coffee, does he? (negative sentential negation)

d. *He doesn’t like coffee, doesn’t he?

2. When we coordinate two sentences and the second one is negative, the

polarity of the first sentence determines the choice of the conjoining element: negative

sentences use and… either and and neither…, while affirmative sentences can only be

followed by and so…:

(4) a. He dislikes coffee and so does his wife. (affirmative sub-clausal negation)

b. *He dislikes coffee and his wife dislikes it either.

c. *He dislikes coffee and neither does his wife.

d. He doesn’t like coffee and his wife doesn’t like it either.

(negative sentential negation)

14

e. He doesn’t like coffee and neither does his wife. (sentential negation)

f. *He doesn’t like coffee and so does his wife.

3. Only negative sentences can take a not even continuation:

(5) a. He doesn’t like coffee, not even cappuccino.

(negative sentential negation)

b. He dislikes coffee, even cappuccino. (affirmative sub-clausal negation)

c. *He dislikes coffee, not even cappuccino.

Due to their proved reliability, these tests can be employed whenever the

negative status of a sentence is questionable, in order to mark the difference between

local negation and negation that takes scope over the entire clause.

2.2. Polarity and Polarity Items

The discussion on polarity items starts from the very trivial observation that, in

any language, there are certain linguistic items (words, phrases or idioms) that can

only appear in negative contexts and are ruled out from affirmative ones. They are

called Negative Polarity Items (NPIs)6. This applies to both grammatical (as in (6)

and (7) below) and lexical (8 – 9) items7:

(6) a. She hasn’t finished writing her paper yet.

b. *She has finished writing her paper yet.

c. She has already finished writing her paper.

(7) a. Nu a scris decât treizeci de pagini.

NOT has written only-NPI thirty pages

‘She didn’t write more than thirty pages.’

6 Huddleston & Pullum (2002: 823) name them ‘negatively-oriented polarity-sensitive items’. Becausethey can appear in other non-assertive contexts like interrogatives, conditionals, comparatives orrelative clauses which modify a superlative (Avram 2009), we could call them Non-Assertive PolarityItems. Apud Zeijlstra (2004: 41), Giannakidou (1997, 1999) argues that the important feature of thecontext in which such items appear is non-veridicality and calls them Affected Items. Just like Zeijlstra,I will, however, keep the traditional name of Negative Polarity Items.7 The lexical vs. grammatical distinction appears in Cornilescu (1982: 8-9)

15

b. *A scris decât treizeci de pagini.

has written only-NPI thirty pages

c. A scris doar treizeci de pagini.

has written only-API thirty pages

‘She wrote only thirty pages.’

(8) a. Joe hasn’t lifted a finger to help8.

b. *Joe has lifted a finger to help.

(9) a. Nu era nici picior de student în facultate.

NOT was NO leg of student in department building

‘There was no student whatsoever in the Department building.’

b. *Era picior de student în facultate.

The examples above show that, for the grammatical items, there are equivalent

terms which are used in the affirmative sentences (already, doar), while the lexical

idioms lack such correspondents. In (8) and (9), the b. sentences can only be

grammatical, as Huddleston & Pullum (2002: 824) explain, if we accept the rather

bizarre literal meaning: Joe’s help actually consisted in lifting one finger and there

was the disembodied leg of a student in the Department building. There are many

such polarity-restricted idioms, and they are generally language-specific9. For

comprehensive lists of such items in English, see Cornilescu (1982: 9) and

Huddleston & Pullum (2002: 823).

On the other hand, the distribution of the grammatical NPIs and their

Affirmative Polarity Item (API) counterparts is more systematical. For English, I

include the table given in Huddleston & Pullum (2002: 831):

Table 3: Polarity Items in English and their correspondence

APIs NPIs Absolute negatorsi. a. Some any no b. someone / somebody anyone / anybody no one / nobody c. something anything nothing d. somewhere / someplace anywhere / anyplace nowhere / no place e. sometimes ever never f. sometime, once anytime, ever never

8 Example (8 a, b) from Huddleston & Pullum (2002: 822)9 For a discussion of several advanced (mainly semantic) topics in the study of NPIs and their licensingconditions, see Zeijlstra (2004: 40-45).

16

g. somewhat at all ---ii. a. still any more / any longer no more / no longer b. Already yet ---iii. a. So --- neither / nor b. too / as well either --- c. --- either neither d. --- either… or neither… nor

2.3. Negative elements

I have mentioned negative contexts, but I did not define them so far. Without

going into any more details, I suggest the following definitions:

(10). a. A negative context is the syntactic environment of an utterance where the

presence of one or more negative elements yields sentential negation.

b. A negative element is a linguistic item which introduces a negative context.

But in every language, there are several types of such constituents that are

used in order to express negation. They can belong either to the lexical domain (like,

for instance, words that have a certain negative meaning in themselves, such as

‘unhappy’ or ‘without’, where we are dealing with affixal negation, or negative

quantifiers like ‘never’ or ‘nobody’) or to the functional domain. These negative

elements have been, in time, variously categorized, but, for the purpose of this paper, I

will adopt the classification used by Zeijlstra in his thesis. He identifies four major

types of negative elements, and summarizes them in the following table (taken from

Zeijlstra 2004: 39):

Table 4:

17

2.4. Interpreting multiple negative elements

This sub-chapter is dedicated to explaining how a sentence with two or more

negative elements (generally a negative marker and one or more n-words) can be

interpreted. There are, generally speaking, two possible interpretations of a sentence

where there are two negative elements: they either cancel each other out (and thus the

sentence is interpreted as affirmative) or they complete each other semantically – and

then the sentence is considered negative. The distinction can be formulated in terms of

the application, or lack of application, of a principle of logics called the Law of

Double Negation (LDN), formulated as follows: Duplex negatio est affirmatio. (“A

double negation is an affirmation”), or, in terms of logical operators:

(11) LDN: ~ (~ p) ↔ p

Languages differ with respect to how they interact with this rule of logics. A

small group of languages (among others, English, Classical Latin, Standard German,

Standard Dutch and the Scandinavian languages) always obey the Law of Double

Negation. For the vast majority, however, logics and linguistics do not converge, and

the languages do not obey the Law.

If in a language the LDN is applied, that language is said to be a Double

Negation10 language. If, however, LDN is not applied, we are dealing with a

phenomenon called Negative Concord, which takes place in “cases where multiple

occurrences of phonologically negative constituents express a single negation” (de

Swart & Sag 2002: 373). Zeijlstra (2004) sees this as a syntactic / semantic distinction

and proposes that Double Negation languages have a semantic negation, whereas

Negative Concord is a type of syntactic negation.

In the case of semantic negation, for each negative operator, there is a

corresponding negative element, in a total, 1:1 equivalence. These elements are

believed to be negative per se, being generated in this way at the level of the lexicon.

Syntactic negation implies that the negative elements only have the role of marking

10 The term can be confusing: we have to distinguish between double negation as a logical andlinguistic phenomenon and Double Negation as a class of languages, especially since the phenomenonoccurs even in languages that are not DN languages. For example, Rom. Nu pot să cred că nu va reuşi.(‘I cannot believe she will not succeed’), meaning Cred că va reuşi. (‘I believe she will succeed’) is aninstance of double negation (as a phenomenon) in a language that is not a Double Negation language.In order to be clear, I will try to refer to Double Negation languages as Non-Negative Concordlanguages whenever possible.

18

the presence of the negative operator, but are not negative themselves, from a

semantic point of view (Zeijlstra 2004: 244). The syntactic character of negative

concord seems to have been deduced, if not formulated into a theory, by older

approaches, which explain the fact that one trait appears in more than one place by

decomposing a [+NEG] element into [neg1] ^ [neg2] (Manoliu Manea 1977: 252).

2.5. A Short History of Double Negation

After having defined Double Negation, it is perhaps appropriate to assume, for

the moment, a diachronic point of view and try to come up with an answer as to how

exactly Latin and English came to be non-Negative Concord languages. Explaining

the evolution of this phenomenon in the two languages will be useful for the purpose

of my present analysis, since I will later argue that the lack of Negative Concord is the

negation type that offers the most clues with regard to the position of the NegP.11

It is a well-known fact that Old English used to allow negative concord and

that, in Romance languages, its usage is often necessary. We are thus faced with a

mirror-like evolution: on the one hand, a language eliminates the negative concord

which existed in its earlier stages; on the other, a mother-language with double

negation gives rise to daughter-languages that admit or even require negative concord.

Whether or not this is a case of linguistic coincidence, I will try to discuss in what

follows. Therefore, the purpose of this chapter is to answer the question: “How come,

despite their contrasting evolution, both English and Latin are Non-Negative Concord

languages?”

In the case of Latin, we could say that negation is “broadened” – from a single

negative element, we reach the point where we may (and must, in fact) use negation in

other places as well, being sometimes forced to do so by the rules of grammar.

Sentences like:

(12) a. Nemo facit quidquam.

NOT-ONE do-PRES 3RD SG NPI-THING

‘Nobody does anything.’

11 This section of my thesis includes material from a short paper (again supervised by prof. L. Avram),which I presented in May 2009 at Conferinţa Cercurilor Ştiinţifice Studenţeşti (The Students’Conference of the Faculty of Foreign Languages and Literatures). It was named Aspects of negation inEnglish and Latin.

19

(13) a. Nemo it usquam.

NOT-ONE go-PRES 3RD SG NPI-PLACE

‘Nobody goes anywhere.’

get to be expressed like:

(12) b. Nimeni nu face nimic. (Romanian)

NOT-ONE NEG do-PRES 3RD SG NOT-THING

c. - Nesunno fa niente. (Italian)

NOT-ONE do-PRES 3RD SG NOT-THING

- Non fa niente nesunno. (Italian)

NEG do-PRES 3RD SG NOT-THING NOT-ONE

d. Nadie hace nada. (Spanish)

NOT-ONE do-PRES 3RD SG NOT-THING

e. Ninguem não faz nada. (Portuguese)

NOT-ONE NEG do-PRES 3RD SG NOT-THING

(13) b. Nimeni nu merge nicăieri. (Romanian)

NOT-ONE NEG go-PRES 3RD SG NOT-PLACE

c. - Nesunno va da nesunna parte. (Italian)

NOT-ONE go-PRES 3RD SG to NOT-THING place

- Non va da nesunna parte nesunno. (Italian)

NEG go-PRES 3RD SG to NOT-ONE place NOT-THING

d. Nadie va a ningun lado. (Spanish)

NOT-ONE go-PRES 3RD SG to NOT-THING place

e. Ninguem não vai a lugar nenhum. (Portuguese)

NOT-ONE NEG go-PRES 3RD SG to place NOT-THING

The data was obtained directly from native speakers through a small online

experiment, in which the members of a forum were asked to translate into their

mother tongues (Italian, Spanish and Portuguese) the two sentences in (12 a) and (13

20

a), given in English, without the Latin equivalents.12 We can observe that,

paradoxically, the structure of the sentences is more similar in Latin and English than

it is between Latin and any of the Romance languages.

If this is a case of “broadening” of negation, in the history of the English

language there was a kind of “narrowing” of negation. In Old English, negative

concord was much more frequent, which can be proven by examples like:

(14) forÞæmple hie hiora nan wuht ongietan ne meahton

because they of them NOT-thing understand NOT could

‘because they couldn’t understand anything of them.’13

(15) Hit nā ne fēoll.

it NOT at all fall

‘It did not fall (at all).’14

(16) Ne maeg ēower nān flæscmetta brūcan būtan mīnun cræfte.

NOT my you NOT-one meat enjoy except for my skill

‘You cannot enjoy meat unless it was for my skill.’15

The process is a gradual one. In Old English, the negative marker ne was

weakened, and, by the Middle English period, it had the form of a preverbal particle

ne / no / na. As it was no longer felt as strong enough, negation had to be frequently

reinforced by means of a post-verbal adverb such as nawiht ‘no creature’ (van

Geldern 2006: 130) or noÞing ‘nothing’:

(17) for of al strengðe ne drede we nawiht

because of all his strength NOT dread we NOT-thing

(18) Never him nas wers for noÞing

‘He was never more distressed about anything.’16

12 The conversations took place between 7 and 9 May 2009 and can still be seen at<http://edufire.com/forums/7-languages/topics/3764-quick-help-please-italian-spanish-portuguese?>13 Pastoral Care, 4/12 in van Gelderen (2006: 70)14 from Smith (2009: 85)15 A Colloquy on the Occupations, King Ælfric, in Mitchell & Robinson 1996: 183)

21

Other cases of Negative Concord in Middle English include examples with

multiple occurrences of the negative marker ne:

(19) I ne can ne I ne mai.

‘I do not know how to nor am I able to.’17

Such multiple negatives will only be lost in late Middle English (van Geldern

2006: 130), leading to the lack of Negative Concord in contemporary English.

Hence, starting from opposing situations, two different languages reach the

same result. There are, however, a few remarks that need to be pointed out regarding

these phenomena. First, a major difference is that the “broadening of negation”

happens with the transformation of one language into another; we are dealing with

distinct languages. For English, the linguistic change from Negative Concord to

Double Negation takes place within one and the same language, but in two different

stages of its evolution. On top of that, the period in which the change was made is

different, with a longer time in the case of Romance negation.

Between the two directions of the evolution, the one undergone by Latin in its

metamorphosis into the Romance languages is easier to explain and seems to be more

logical. The speakers’ need for expressive force can introduce new elements into the

language, especially for such an important phenomenon. This is the main argument

that Jespersen (1917: 5) uses to explain the change18. Negation is a formally weak

element, but its notional (i.e., semantic) role is immense, perhaps greater than that of

any other logical operator in a sentence. Hence, the speakers are naturally inclined to

add something more, to include a morphologically negative element, in order to give

negation a more powerful form and make sure the listener understands them:

“It might, however, finally be said that it requires greater mental

energy to content oneself with one negative, which has to be remembered

during the whole length of the utterance both by the speaker and by the hearer,

16 in Burrow & Turville-Petre (1996: 53)17 in Burrow & Turville-Petre (1996: 52)18 The theory of the diachronic development of Negation is called the Jesperson Cycle and issummarized in Zeijlstra (2004: 56-57), with an application on negation from Latin to ModernColloquial French.

22

than to repeat the negative idea (and have it repeated) whenever an occasion

offers itself.” (Jespersen 1917: 72)

This very commonsensical argument can account for why exactly the

Jespersen Cycle is put into motion in the first place. It happens especially in languages

where negation is expressed by a phonetically reduced element. According to

Jespersen, the first step in strengthening negation is the grammaticalization of certain

lexical elements and their transformation into markers of negation. When these, too,

cease to be enough, language evolves towards Negative Concord.

This explanation may be true for the evolution of negation in Romance

languages but, if we are to admit that this is the normal tendency, how come a

language like English suffered the reverse phenomenon? At first glance, this is indeed

a paradox.

The answer starts from the fact that, so far, in everything that I said, I have

willingly ignored an essential aspect: the claim that Latin and English are Double

Negation languages is sustainable only for a restricted sector of the two languages,

which is constrained by both temporal and register factors. What I called “Latin” in

this paper is actually nothing else but the literary, written Latin of the Classical

period, that is, used in a short interval (1st c. B.C. to 1st c. A.D.) and belonging to the

higher register. Also, the term “English” was used exclusively to refer to Modern

Standard English, that is, the most polished form of the language. Therefore, the

Double Negation that the two languages exhibit could be the result of a

standardization process.

In fact, the other registers of Latin and English exemplify many cases of

Negative Concord. We are dealing with situations where there is no ambiguity and a

number of negative elements result in a negative interpretation of the sentence. I will

offer some examples taken, for Latin, from the literary works that aim to imitate the

colloquial spoken language and, for English, from the lower registers.

The Roman poet Plautus provided numerous examples of Negative Concord,

since his is a language equally pre-classical, oral and colloquial. On top of that,

comedy is a literary genre where the language is chosen to be as expressive and

colorful as possible. All the examples below are taken from the Latin Syntax of

Touratier (1994), but the translations are mine.

23

(20) Iura te non nociturus esse homini (…) nemini!

swear-IMP you-ACC NEG harmful be-INF man-DAT NOT-ONE-DAT

‘Swear that you will not harm anyone!’

(21) Nec meus servus numquam tale fecit quale tu mihi.

NOT-AND my slave NOT-EVER so do-PERF 3RD SG as you me-DAT

‘No slave of mine ever did what you did to me.’

Examples of colloquial and expressive language, this time from the Imperial

(post-classical) period can be found in the speech of Petronius’ characters in his novel

The Satyricon:

(22) Neminem nihil boni facere oportet.

Not-ONE-ACC NOT-THING good-GEN do-INF should

‘Nobody should do any good.’

We could also add phrases like neminem nihil („nothing no-one”) and nemini

tamen nihil satis est („nobody is satisfied with anything”), which the editor of the

Oxford edition mentions. The last example, just like (20) and (21), with the verb in

the affirmative form and two negative elements in the lexical domains, show that the

type of Negative Concord in Latin was different from that of Romanian and closer to

that of Spanish or Italian. But it is not only the speech of the lower classes of society

that included Negative Concord. In the pre-classical period we also find it in Ennius,

who was endeavoring to use the high style of tragedy:

(23) Lapideo sunt corde multi, quos non miseret neminis.

who-ACC NEG produce mercy-IND NOT-ONE-GEN

‘Many have stoned hearts – they don’t feel pity for anyone.’

The most surprising example, however, comes from an author whom many

consider “the most classical” of them all: Cicero himself, who, in his oration against

Verres, states:

24

(24) Decebat Epicrates nummum nullum nemini.

owe-PERF 3RD SG Epicrates money-ACC NOT-THING-ACC NOT-ONE-DAT

‘Epicrates didn’t owe anyone any money.’

In this case, modern editors do not correct nullum to ullum (the corresponding

Negative Polarity Item), as opposed to another example of Ciceronian Negative

Concord, in his Philipicae, where the manuscript rendition of an ambiguous nullam is

generally accepted as an ullam. (Touratier 1994: 473). The context in which the

sentence in (24) was pronounced was far from being either colloquial or vulgar – we

have in fact a sample of Negative Concord in the cultivated register, justified,

perhaps, by the expressive effects of alliteration.

From the examples given so far, we can see that Latin only had Double

Negation in its most polished aspect, that of written, Classical Literary Latin. But the

Romance languages were not formed from this register, but the vulgar, colloquial,

spoken one. Here is, therefore, our first answer: the Romance languages evolved

Negative Concord by expanding the concord tendencies that already existed in Vulgar

Latin. This hypothesis is in agreement with Jespersen’s considerations that I have

mentioned above.

As for the restricting of negation in Modern Standard English, we have to say

that this is not in fact an evolution, but a process of selection: from a number of

variants there is one chosen, and this one is later imposed as the standard.

In fact, rare instances of Negative Concord can still be not only found in

Shakespeare’s Early Modern English:

(25) Nor go neither:… and yet say nothing neither. (The Tempest)19

but also in the Modern English of such great Victorian novelists, when they want to

render the speech of less educated characters:

(26) ‘Nobody never went and hinted no such thing,’ said Peggotty. (Dickens)

19 in van Gelderen (2006: 172)

25

(27) All he [the butler] hopes is, he may never hear of no foreigner never boning

nothing out of no travelling chariot.’ (Dickens)

(28) We never thought of nothing wrong. (Thackeray)

(29) There was niver nobody else gen (gave) me nothin’. (George Eliot)20

Apart from these literary examples, many non-standard varieties of English

display Negative Concord. Here are a few authentic sentences which exemplify

Negative Concord in contemporary British (regional) English. They were spoken by

the speakers with the purpose of expressing a single negation and it is not possible to

interpret them as Double Negation:

(30) I’m not never going to do nowt more for thee.

‘I’m not ever going to do anything more for you.’

(31) He couldn’t find none nowhere.

‘He couldn’t find any anywhere.’

(32) I never saw nothing.

‘I didn’t see anything.’21

Such sentences would have been ruled out in Standard English, and their

speakers catalogued as uneducated, because of what Jespersen (1917: 65-66) calls “a

tendency ridiculed at school, however natural in itself”. The same idea is expressed,

more clearly and more recently, by Huddleston & Pullum as well:

“There is an extremely widespread tendency among Standard English

speakers to regard dialects with negative concord as ‘illogical’ or ‘inferior’. It

is argued that by a rule of logic two negatives cancel each other out to make a

positive22. (…) But such an argument is completely invalid. The rule of logic

20 Examples (26) – (29) from Jespersen (1917: 66)21 Examples (30) – (32) from Avram (2009)22 the Law of Double Negation stated in (11) above.

26

that two negatives are equivalent to a positive applies to logical forms, not to

grammatical forms. It applies to semantic negation, not to the grammatical

markers of negation. (…) The difference (…) is a matter of grammar, not

logic, and neither set can be regarded as intrinsically superior to the other.”

(Huddleston & Pullum 2002: 847)

Therefore, if Negative Concord was present in Old English and continues to

exist in the majority of sub-standard dialects of present-day English, and, what is

more, this is the natural evolution, then what was the reason negation was restricted to

a single element in Standard English?

The answer can be found in the history of English. As Huddleston & Pullum

(2002: 847) explain, the decline of Negative Concord “may have had much to do with

a nascent prescriptive tradition and its conscious comparison of English and Latin.”

The authors of these prescriptive grammars even acknowledged the phenomenon. The

most influential of these grammars were that of Robert Lowth in 1710 and that of

Lindley Murray in 1795 (Crystal 2004: 396). Murray’s grammar states that “in

English, two negations destroy one another, or have the meaning of an affirmation”.

This grammatical rule is influenced, on the one hand, by the relations holding between

mathematical and logical operators and, on the other… by the existence of Double

Negation in Latin.

2.6. Conclusions

This chapter was a very short presentation of some of the most important

issues in the study of negation. First of all, we distinguished three types of linguistic

negation: two sub-clausal types (word / affixal negation and phrasal / constituent

negation) and a clausal one – sentential negation.

In order to prove that, in a particular utterance, negation has sentential scope,

one can use the diagnosis tests elaborated by Klima (1964), that we have reviewed

here: the neither / either test, the not – even test and the test involving the use of tag

questions with complementary polarity.

Then, I mentioned and exemplified Polarity Items, giving a detailed list of

grammatical NPIs in English, together with their Affirmative Polarity Items

counterparts and the absolute negators they can replace.

27

The next step in describing negation was to adopt a classification of the

various negative elements that can be found in a sentence: they are negative

markers, n-words, negative quantifiers and semi-negatives.

If more than one such negative element appears in a context, we showed that

the interpretation may be different according to the type of negation that particular

language displays: if it is a Double Negation language, the Law of Double Negation

will be applied, the two negators will cancel each other out and the meaning of the

sentence will be perceived as affirmative; if, on the other hand, the language has

Negative Concord, the Law of Double Negation will not be taken into consideration

and the sentence will be interpreted as being negative.

This is considered to be the normal interpretation in natural languages, and, in

the last sub-section, I have exemplified the evolution from Double Negation to

Negative Concord in the Romance languages. For English, the reverse evolution is

observed, due to the influence of prescriptive grammars that tried (and, in fact,

managed) to model English after logics and Latin – a tendency that is twice artificial:

firstly because the imposed linguistic system was unrelated, and secondly because this

very system was, itself, artificially created in order to obey the laws of logics.

After all these general considerations, I will try, in the next chapter, to apply

the terminology and classifications established so far to the description of negation in

Classical Latin.

3. An overview of negation in Latin

This chapter deals with the nature and distribution of Latin negators, with

special attention given to those “link words” used in coordination and subordination. I

start by presenting the various types of negative elements in Latin (3.1), offering a

more in-depth analysis of negative quantifiers and their use in relation to the non-

Negative Concord status of Latin (3.1.4). Then, in 3.2, I present the distribution of

Negative Polarity Items in Latin, which is shown to be very similar to that in English.

Finally, in 3.3, I discuss several cases of Double Negation (with the expected

affirmative reading), while 3.4 is a much-needed summary of this chapter.

28

3.1. Negative Elements in Latin

In the previous chapter (2.3), I adopted a classification of negative elements. In

what follows, I will try to adapt this classification to the situation found in Latin. The

examples, taken from classical authors, were either found by me, or taken from

different grammar books.

3.1.1 Negative markers

There are three separate negative markers in Latin: haud, non and ne.

Haud, as all the grammars agree, is used solely for constituent negation,

usually before an adverb or an adjective, but also before pronouns or nouns, like in

this example (Cic., de Orat.23, II, 26, 101):

(1) Sapiens virtuti honorem praemium, [haud praedam] petit.

wise man virtue-DAT honor reward-ACC, NOT spoil-ACC ask-PRES IND 3RD

‘The wise want to receive honor for their virtue as a reward, not as spoils.’

Haud is rarely used as a verbal negator in Classical Latin, mostly in the fixed

phrases haud scio (‘I don’t know’) and haud dubito (‘I don’t doubt’).

By contrast, non can signify both constituent negation, as in Cic., Att. II, 1, 8:

(2) Nam Catonem nostrum [non tu] amas plus quam ego.

for Cato-ACC our-ACC NOT you love-PRES 2ND more than I-NOM

‘As for our friend Cato, you do not love him more than I do.’ (trans. Evelyn

Shuckburgh)

… and sentential negation, like in the famous quote from Cic. Cat. I. 1. 1:

(3) Patere tua consilia non sentis/, constrictam.. coniurationem tuam non vides/?

‘Don't you feel that your plans are exposed, don't you see that your conspiracy

is already restrained?’

23 For Latin author names and titles of literary works, I have used the common abbreviation systemused in Classical studies.

29

The third negative marker of Latin is ne, which, as a conjunction, is the

negative equivalent of the affirmative ut and introduces subordinates which, as I will

argue later, have a feature of irrealis. The negative conjunction ne must be related to

the affix *n- or possibly *nĕ-, and also, as Sluşanschi (1994: 109) notes, to the enclitic

interrogative particle -ne24, which seems to have evolved from “no?” to a purely

interrogative meaning.

Concerning the etymology of this ne, Jespersen (1917) explains that it comes

from

“…the old negative ne, which I take to be (together with the variant me) a

primitive interjection of disgust, accompanied by the facial gesture of

contracting the muscles of the nose (Dan. rynke på næsesen, G. die nase

rümpfen, Fr. froncer les narines25; the E. ‘to turn, or to screw, up one's nose’ is

not so expressive). This natural origin will account for the fact that negatives

beginning with nasals (n, m) are found in many languages outside the Indo-

European family.” (Jespersen 1917: 6-7)

However, Costa (1999: 93) warns us that, charming as it might be, this

assumption, just like all the hypotheses concerning such an early phase of Indo-

European, should be taken cum grano salis.

I will be returning to both ne and non in the next chapter, where I try to

account for their distribution and positions in the syntactic tree. For now, let us

continue our inventory of Latin negators.

I want to argue that the so-called “negative link words”, that is, negative

copulative conjunctions, have to be included here, in the class of negative markers.

There are two pairs of coordinators: neque and nec, which include a negative meaning

and a copulative meaning (not + and), and neve and neu, which, apart from negation

and coordination, also have the irrealis feature associated with ne (and + no + to).

Etymologically, we find that neque is formed by merging the short ne and the

enclitic copulative particle -que. In the case of neve, there is the long ne, plus the

24 Like, for example in Vidisne?, ‘Do you see?’, which can also be translated by ‘Don’t you see?’.Generally, the illocutionary force of negation and interrogation are, to a point, overlapped, possiblybecause they must occupy the same position within ForceP.25 See also Rom. a strâmba din nas.

30

enclitic -ve, which comes from the disjunctive domain, but receives a copulative

meaning (Ernout-Thomas 1989: 150)26. Nec and neu, respectively, are simple

variants. For the neque / nec distinction, Costa (1999: 93) points out that the

distribution of the two is “independent of any semantic or phonetic considerations”,

being, rather, a variation in register: nec is preferred in the lower strata of Latin, while

the highly cultivated language tends to use neque more often.

Let us now examine the syntactic contexts in which these negative markers,

incorporating a feature for copulative coordination, can appear. The list of syntactic

environments where they can appear is taken from Oudot (1964: 535 onwards).

A. Neque and nec (= et non) are used to link:

a. A negative constituent to an affirmative constituent:

i. In two different sentences, i.e., at the level of the discourse –

the complementizer is used in order to link the sentence to the

larger environment of the text:

(4) Romani… quid petunt aliud… nisi… his aeternam iniungere servitutem?

Neque enim ulla alia conditione bella gesserunt. (Caes. Gal. VII, 77, 15)

‘But what other motive or wish have the Romans, than… to… impose on them

perpetual slavery? For they never have carried on wars on any other terms.’

(trans. W.A. McDevitte)

ii. In two different clauses (within a complex sentence):

(5) Promittis ad cenam nec venis. (Plin. Ep. I, 15, 1)

promise-PRES 1ND to dinner-ACC SG NOT-AND come-PRES 2ND

‘You promise to come to dinner and you don’t come.’

b. Two negative constituents:

i. In two different clauses:

26 Not all scholars agree. For a larger discussion on the various possible etymologies, see Costa (1999)

31

(6) Nihil de sententia dicturus sum… neque hos habendos civium loco… censeo.

NOT-THING NOT-AND

‘I will say nothing about their opinion… and I do not think they should be

considered citizens’ (Caes. Gal, VII, 77, 3)

‘I shall pay no attention to <their> opinion… nor do I think that they ought to

be considered as citizens.’ (trans. W.A. McDevitte)

ii. Two constituents belonging to the same clause, i.e., sub-clausal

constituents:

(7) Erat Romanis nec loco nec numero aequa contentio.

AND-NOT AND-NOT

‘Neither in position nor in numbers was the contest an equal one to the

Romans’ (trans. W.A. McDevitte – Caes. Gal., VII, 48, 4)

Of course, the construction et non can and does appear, either in the case of

constituent negation:

(8) Pugnax et acer et [non rudis] imperator. (Cic., Mur., XV, 32)

‘A belligerent and severe and not inexperienced commander’

(9) Quid tu fecisses, si te Tarentum et [non Samarobrivam] misissem?

‘What would you have done, if I had sent you to Tarentum, and not to

Samarobriva?’ (Cic., Fam., VII, 12, 1)

… or “for pragmatic reasons”, in order to underline an opposition:

(10) Videmus… et non commovemur. (Cic., Har., XII, 25)

‘We see… and we are not impressed’

B. Neve and neu (= et ne)27 are used to link:

27 The equivalence is reinforced in Ernout-Thomas (150), which also gives the archaical forms neiveand nive

32

a. A negative and an affirmative constituent, if they are both clauses:

(11) Obsecrant/ ut suis fortunes consulat/ neu se ab hostibus diripi patiatur/.

entreat-IND 3RD to protect-subj 2nd AND-NOT-TO allow-SUBJ 2ND

‘<They> solemnly entreat him to protect their property, and not to suffer them

to be plundered by the enemy’ (trans. W.A. McDevitte – Caes. Gal., VII, 8, 4)

b. Two negative constituents:

i. If they are both clauses:

(12) Cohortatus est ne se admodum animo demitterent

encouraged-IND 3RD SG NOT-TO sink-SUNJ 3RD PL

neve perturbarentur incommode.

AND-NOT-TO alarm-SUBJ 3RD PL PASS

‘Encouraged his soldiers… that they should not be too much depressed in

spirit, nor alarmed at their loss.’ (trans. W.A. McDevitte – Caes. Gal., VII, 29)

ii. If they are words in the same clause:

(13) Litteras nuntiosque misit, ne eos frumento neve alia re iuvarent.

NOT-TO AND-NOT-TO help-SUBJ 3RD PL

‘sent letters and messengers… (with orders) that they should not assist them

with corn or with any thing else’ (trans. W.A. McDevitte – Caes. Gal., I, 26, 6)

3.1.2 Semi-negatives

Here, the terminology becomes a little confusing, as traditional grammars, at

least French ones, tend to put together n-words, negative quantifiers and, occasionally,

even Negative Polarity Items, under the general term ‘les semi-négatives’. A

definition of those, obviously faulty, describes them as “words (adjectives, pronouns,

adverbs) which, because they are, very often, used together with a negation, preserve

a certain negative meaning even when they are used alone” (Oudot 1964: 109).

33

In Zeijlstra’s (2004) terminology, true semi-negatives are those words that,

without producing any kind of syntactic negation, have a semantically negative

meaning. We could include Latin examples like the preposition sine ‘without’, the

adverb vix ‘scarcely’ or, based on Sluşanschi’s (1994: 28) elucidation on its dubious

behavior with respect to negation, even the verb timeo ‘to fear’.

3.1.3 N-words

As n-words occur only in Negative Concord languages, we can safely assume

that all other negative elements in Latin are negative quantifiers.

3.1.4 Negative quantifiers

Because, just like in English, the single Latin negation is often expressed in the

lexical domain, the negative quantifiers are many and very important. In fact, if we

take into consideration that (1) Latin is a non-Negative Concord language, therefore

negation can only be expressed once, (2) normal word order in Latin usually places

the verb at the end of the clause and (3) there is a general tendency to place negation

on the first available element that can carry it28, then we can predict that, statistically,

sentential negation by means of a negative quantifier has to be much more frequent, in

Latin, than sentential negation with a negative marker29.

The different negative quantifiers in Latin, together with their NPI

counterparts and English translations, are organized in Table 5 below:

Table 5: Latin negative quantifiers and NPIs

28 This is a major point in the analysis developed in the next chapter.29 This, of course, remains to be proved by extensive corpus analyses.

Grammarical

Category

Latin Negative

Quantifier

English equivalent Latin Negative

Polarity Item

English equivalent

Determiner nullus,-a,-um No ullus,-a,-um any

pronoun – person nemo / nullus no-one, nobody quisquam anyone, anybody

pronoun – thing nihil nothing quicqusm anything

adverb – time numquam never umquam ever

adverb – place nusquam nowhere usquam anywhere

34

A few observations are in order. First of all, we should not forget that nullus is

fully inflected for gender, number and case and has the forms of an adjective of the

first class (-us, -a, -um). It can be used either as a determiner, together with a noun

(e.g. nullus vir ‘no man’, nulla femina ‘no woman’) or alone. Also, it is used as the

plural of nemo. Nemo itself has a few inflected forms: N. V. nemo, G. (rarely)

neminis, D. nemini, Acc. neminem. For Ablative and often for Genitive as well, it is

replaced by the corresponding forms of nullus.

In the NPI column, we can see that quisquam and quicquam are forms of the

same indefinite pronoun, with quisquam being the animate (masculine and feminine)

variant, and quicquam the inanimate (neuter) one. It has all the case forms, including

those of the plural. Another indefinite pronoun, aliquis (masc, fem), aliquid (neuter),

functions as the Affirmative Polarity Item, just like the English ‘someone’ /

‘somebody’ and ‘something’.

With respect to their etymology, Ernout-Thomas (1989: 149) show that the

quantifiers for person and thing are the result of a process of grammaticalization: nihil

< *nĕ + hilum and nemo <*nĕ + hemo, where hemo is an ancient form for homo,

‘man’, while hilum is an old noun meaning ‘trifle’, most probably a phonetic variation

of filum ‘thread, string’. The two grammarians draw attention to the difference

between the negative marker ne (which has a long vowel and should have been

represented as nē) and the unattested variant with a short vowel, *nĕ30.

The negative quantifiers for place and time seem, surprisingly, to have been

derived from the Polarity Items, negated with the same element: numquam < *nĕ +

umquam, nusquam < *nĕ + usquam. The same for nullus < *nĕ + ullus.

Before passing on to the section on Negative Polarity Items, I have to mention

the general tendency of favoring a [negative marker + NPI ] construction instead of an

[affirmative conjunction + negative quantifier] one, following the rules that school

teachers always have trouble instilling in their pupils:

Table 6: NPI / negation word order

We don’t say… … but

et nullus neque ullus

et nemo neque quisquam

30 Sluşanschi’s (1994: 105) opinion is that this distinction is irrelevant. After all, nē, *nĕ and even theaffix *-n- can all be formed from the same PIE root.

35

et nihil neque quicquam

et numquam neque umquam

et nusquam neque usquam

The same holds for subordination conjunctions (use ne quisquam or neve

umquam instead of ut nemo and ut numquam), as we will see extensively in the next

chapter. Of course, such constructions did exist and can be found in texts, even in the

Classical period – bun most of them can be shown to be cases of local, constituent

negation, with the scope of the negation limited to the subject / time / place

specification.

3.2. Negative Polarity Items

In Latin, NPI are strikingly similar, in their distribution, to the ones in English.

Again, I follow Oudot (1964: 111 onwards). He, however, calls them semi-negatives.

That is why, wherever the terminology differs, I will be mentioning both names. Here

is, therefore, the distribution of Latin Negative Polarity Items:

A. NPIs in a negative clause:

a. With a “simple negation”, i.e. negation is expressed in the functional

domain, by means of a negative marker:

i. Haud:

(14) Patrum haud fere quisquam in foro. (Liv., III, 38, 11)

Senator-GEN PL NOT almost any in forum

‘Almost none of the senators was in the forum.’

ii. Non:

(15) Si non ulla tibi facta est iniuria (Cic., Div. Caec., 18. 60)

‘If there wasn’t any injustice done to you’

36

(16) Non umquam turpior in ludo talario consessus fuit. (Cic., Att., I.16. 3)

NOT ever more disgusting in game of dice assembly was

‘There never was a seedier lot round a table in a gambling hall.’ (trans. Evelyn

Shuckburgh)

b. With a “negative link word” (liaison négative), i.e. negative

complementizer – nec, neque, neve, neu:

(17) …nec quicquam iam supererat… (Liv, III, 37, 4)

NOT-AND anything already remain

‘And already there didn’t remain anything….’

(18) Veni Athenas… neque me quisquam ibi adgnovit. (Cic., Tusc., V, 104)

NOT-AND me anyone there recognized

‘I came to Athens... and nobody recognized me there.’

c. With a “compound negation” (negation composé), i.e., negation is

expressed in the lexical domain, by means of a negative quantifier:

(19) Nihil cuiquam sanctum (Liv., II, 55, 8)

NOT-THING anybody-DAT sacred

‘Nothing sacred to anyone’

B. NPIs with semi-negatives31:

a. With the preposition sine ‘without’:

(20) sine ullo metu (Liv., II, 33, 6)

‘without any fear’

31 Here my classification differs drastically from Oudot’s – he places semi-negative contexts under thesame heading with negative marker environments.

37

b. With a verb with a negative meaning (nego ‘to deny’, nolo ‘not to

want’, caveo ‘to beware’, nescio ‘not to know’, veto ‘to forbid’, desino

‘to cease’, etc.):

(21) Vetat quemquam privatae quicquam rei agere. (Liv., III, 27, 2)

anyone-ACC SG anything-AC SG

‘forbade anybody to engage in any private business.’ (trans. B. O. Foster)

(22) Bibliothecam tuam cave cuiquam despondeas. (Cic., Att., I, 10, 4)

library-ACC SG your avoid-IMP anyone-DAT SG promise-SUBJ 2SG

‘Mind you don't engage your library to anyone’ (trans. Evelyn Shuckburgh)

C. Without a negative word, i.e., in affirmative contexts32:

a. In interrogatives and exclamatives:

(23) An est quisquam qui hoc ignoret? (Cic., Mil., 3. 8)

INT PART is anyone-NOM SG who this ignore-SUBJ 1ST

‘Is there any one who does not know…’ (trans. C.D. Yonge)

(24) Quando… Socrates… quicquam tale fecit? (Cic., Fin., II, 1, 1)

when anything-ACC SG such do-PERF 1ST

‘…when did Socrates, … do anything of the sort?’ (trans. C.D. Yonge)

(25) Hic mihi quisquam mansuetudinem et misericordiam nominat!

here to me anyone gentleness-ACC and compassion-ACC name-SUBJ 1ST

‘In such a case, does any one talk to me of gentleness and compassion?’ (trans.

John Selby Watson – Sal., Cat., LII, 11)

b. In conditionals:

(26) Si quisquam, ille sapiens fuit. (Cic., Amic., II, 9)

‘If anyone, that one was wise’

32 In fact, what the following sub-headings have in common (among themselves and in common withnegation) is a non-assertive feature.

38

c. In comparatives:

(27) Refertius erit aerarium… quam umquam fuit. (Cic., Ver., II, 3, 202)

crammed-COMP be-FUT treasury than ever was

‘The treasury will be more crammed than it ever was!’

A certain remark has to be made: if the situations described in C above are

supposed to be negative, the negative word is used (Oudot 1964: 114). Thus, we have:

a. Negative conditionals:

(28) Si nihil erit praeter ipsorum suffragium, tenue est.

if NOT-THING be-FUT 3RD beyond their vote little is

‘If they have nothing beyond their own vote, that is but little’ (trans. C.D.

Yonge – Cic., Mur., XXXIV, 71)

b. Negative interrogatives:

(29) An mihi de te nihil esse dictum umquam putas?

INT PART me-DAT about you NOT-THING be-INF said ever think-PRES 2ND

‘Do you think that I haven’t heard anything about you?’ (Cic., Fam. III, 8, 6)

3.3. Double Negation

After having examined, in turn, the distribution of Latin negative coordinators,

negative quantifiers and NPIs, we cannot finish our overview without a brief

presentation of the mechanics of Double Negation. In the previous chapter, I have

provided both a working definition for it, and a short evolution of Latin (and English)

lack of Negative Concord. Now it is time to offer a few concrete examples of

sentences where the speaker intentionally uses two negative elements with the exact

purpose of expressing an affirmation. Given the relative sophistication of Double

Negation, which requires not only an intricate grammatical structure, but also a

certain complexity in thinking, we generally find such constructions in texts dealing

39

with philosophical themes. The two negations will cancel each other out, as expected

in the domain of logical and mathematical operators.

The most frequent type of Double Negation is that in which a negative verb in

the infinitive is negated a second time by the sentential negation that is placed either

on the verb, such as in

(30) Iuvenem nostrum non possum [non amare]. (Cic., Att., X, 10, 6)

youth-ACC our-ACC NOT can-PRES IND 1ST NOT love-INF

‘I can’t not love our young man.’ i.e., ‘I have no choice but to love him.’

… or on a negative quantifier:

(31) Si enim sunt, nusquam esse non possunt. (Cic., Tusc. I, 6, 11)

NOT-PLACE be-INF NOT can-PRES IND 3RD

‘So, if they exist, they cannot be nowhere.’, i.e.,

‘For, if they exist at all, they must be somewhere.’ (trans. C. D. Yonge)

(32) Nihil relinquere non est ausus. (Cic., Fin., IV. 16. 43)

NOT-THING leave-INF NOT dare-PERF IND 3RD

‘He did not dare not to leave it behind.’ i.e., ‘He left it behind, as he did not

dare to take it.’

As clearly shown, the first (almost word-for-word) translations of the above

examples are perfectly correct and interpretable English sentences, because the same

mechanism is at work. But such constructions exist in Negative Concord languages as

well, and, because, in these languages, the negative elements form a unique negation,

it is more difficult, in some case, to create the desired affirmative interpretation. I will

try to find equivalents in Romanian for each of the examples in this section and

compare the structures.

So far, the situation is simple enough – because the interaction of the two

negations takes place at the level of the complex sentence, we have an identical

construction in Romanian, with a negated matrix and a negated subordinate clause

that yield an affirmative reading:

40

(30 a) Nu pot [să nu-l iubesc].

(31 a) Nu a îndrăznit [să nu îl părăsească].

(32 a) Dacă există, ei nu pot [să nu fie niciunde].

(32 b) ?Dacă există, ei nu pot [să fie niciunde].

For the data in (32), however, the interpretation of the translations is more

difficult. Notice that *Ei pot să fie niciunde would be clearly ungrammatical, but, in

the context in (32 b), the construction is, I think, acceptable.

Yet more complicated is the situation in which the negating-a-negation-

results-in-an-affirmation interpretation is obtained by inserting a constituent negation

on the verb within a clause sententially negated by a negative marker, as in

(33) Nec hoc ille [non vidit]. (Cic., Fin., IV, 22, 60)

NOT-AND that he NOT see-PERF IND 3RD

‘And he didn't [not see it]’ i.e., ‘And it is not the case that [he did not see it].’

… which is practically impossible to translate into Romanian.

Finally, I will conclude the section on Double Negation with the analysis of a

thought expressing the idea of universal unhappiness: Nemo non miser. The phrase

appears several times in the corpus, among others, in these two very clear examples

from Cicero:

(34) ‘Nec vero quisquam stultus non horum morborum aliquo laborat,

NOT-AND NOT suffer-PRES IND

nemo igitur est [non miser].’ (Cic., Fin., I, 59)

NOT-PERSON therefore is NOT miserable

'Nor is there any fool who does not suffer under some one of these diseases.

Therefore there is no fool who is not miserable.' (trans. C. D. Yonge)

Here, the unlikely appearance of the copula signals that the negation is to be

understood as taking scope on the adjective only, not even on the whole Small Clause.

41

But in the next example, because of the absence of the verb ‘to be’, the interpretation

is ambiguous:

(35) M.: Ergo et ii, quibus evenit iam ut morerentur, et ii, quibus eventurum est,

miseri?

A.: Mihi ita videtur.

M.: Nemo ergo non miser?

A.: Prorsus nemo. (Cic., Tusc., I, 5, 9)

‘M. “Then those who have already died, and those who have still got to die,

are both miserable?”

‘A. “So it appears to me.”

‘M. “Then all are miserable?”

‘A. “Every one.” ’ (trans. C. D. Yonge)

‘Then is there no-one who is not miserable’ // ‘Exactly, no-one.’

The double negation in (35) can be perceived to be either (35 a) or (35 b)

below, where (35 a) is the reading that holds in (34) as well (Nemo est [non miser]),

and (35 b) is the reading of a sentence with negation taking scope over the whole verb

phrase (Nemo [non est miser]):

(35 a) ~ ( x) (x is (~ miserable)

(35 b) ~ ( x) ~ (x is miserable)

Touratier (1994: 468-469) compares the interpretation of (35 b) with a possible

sentence (35 c):

(35 c) Non [nemo est miser].

~ ( ~ ( x) (x is miserable) )

He also draws two syntactic trees (Touratier 1994: 468), which I will copy

here – even if, by current generative grammar standards, these representations are out-

42

dated, they are very clear33 and helpful in identifying scope relations.

(36) P P┌────┴────┐ ┌────┴────┐

Neg P SN SV┌───┴───┐ ┌───┴───┐

SN SV V Adj┌─┴─┐ ┌─┴─┐

V Adj Neg V

Non nemo est miser Nemo non est miser

Hopefully, after elaborating our approach on the locus of the negative

elements, I will be able to draw proper syntactic representations for these sentences.

3.4. Conclusions

This chapter was a short overview on negation in Latin. We suggested an

inventory of Latin negators, including: (1) the negative markers haud, non, ne and

negative conjunctions, neque/nec and neve/neu, (2) semi-negatives that licence NPIs

(sine, caveo, veto etc) and (3) negative quantifiers for person (nemo, nullus), thing

(nihil), place (nusquam) and time (numquam), with their corresponding Negative

Polarity Items. Finally, we have seen a few examples that illustrate how Double

Negation functions in Latin.

In the final chapter of my diploma paper, I will try to bring together everything

that I have included so far and tentatively suggest a hypothesis concerning the place

of negative elements within the fine-grained structure of the split CP.

4. Arguments in favor of Negation at the Left Periphery

This chapter consists of several seemingly non-related arguments that I will

bring into discussion in order to support my analysis of negative complementizers. In

the end, the argumentation will be shown to converge.

33 I keep the French abbreviations: P = proposition, SN = syntagme nominale (=NP), SV = syntagmeverbale (=VP).

43

4.1. Scopal Ambiguities

In this subsection, I will offer several arguments supporting the idea that

negation is a linguistic phenomenon whose semantic complexity can perhaps excel the

explanatory powers of syntax.

The first problem that comes to mind is the observation that sometimes

negation can appear in a certain place in a sentence, but be understood to take scope

over something else. As Moscati (2006: 10) points out, “there are cases, in fact, where

negation might be interpreted in a position different from the one in which it is base-

generated, suggesting that the relation between lexical insertion and interpretation is

not as straightforward as it was once thought”. He offers34 the following two

examples:

(1) a. I don’t think that he has come.

b. I don’t cut my salami with a hacksaw.

(1) a. is a classical case of Negative Raising, defined as “a lexically governed

rule that transports the negative constituent (…) from an embedded complement

clause into the main clause” (Cornilescu 1982: 34). We can see the negative marker

on the verb in the matrix, but we understood it as belonging to the predicate in the

direct object clause, as in (2) a. below.

(2) a. I think that he hasn’t come.

b. I cut my salami not with a hacksaw (, but with a knife).

(2) b. is, likewise, a rephrasing of the (1) b. sentence, with the negative marker

placed directly above the constituent which we understand as negated. Note that in (1)

b. and (2) b. the syntactic scope of negation in different: (1) b. has sentential negation,

while (2) b. has constituent negation, as demonstrated by applying any of Klima’s

(1964) tests:

(3) a. I don’t cut my salami with a hacksaw, not even when I’m terribly hungry.

34 First example from Jespersen (1917), second from Stockwell, Schachter, Partee (1973).

44

b. *I cut my salami not with a hacksaw, not even when I’m terribly hungry.

c. I don’t cut my salami with a hacksaw and neither does the blacksmith.

d. I cut my salami not with a hacksaw, and so does the blacksmith.

However, despite the fact that negation in (1) b. is sentential and syntactically

dominates the entire clause, only a part of the sentence is really negated: I do indeed

cut my salami, but I do not perform the cutting operation with a hacksaw. By

displacing the negative operator, a verb that is actually affirmative in its meaning is

rendered syntactically negative.

It is perhaps interesting to point out that this kind of ambiguity is in no way

related to the fact that English is a language that only accepts one negative element in

a clause. Even in Negative Concord languages, the situation is the same, as illustrated

for Romanian below:

(4) a. Nu îmi tai salamul cu bomfaierul.

b. *Nu îmi tai salamul nu cu bomfaierul.

c. Îmi tai salamul nu cu bomfaierul (, ci cu cuţitul).

The fact that sentential negation can take semantic scope over any of the

elements in a clause was noticed by Huddleston & Pullum (2002: 790-1). They offer

the example in (5) and argue that, in order for (5) a. to be true, all the statements in (5)

a. i. – iii. also have to be true:

(5) a. Liz deleted the backup file.

i. “A deletion operation took place.”

ii. “The deletion operation was performed by Liz.”

iii. “The deletion operation was performed on the backup file.”

b. Liz didn’t delete the backup file.

(5) a. will automatically become false if any single one of the three condition

in i. – iii. are false, as (5) b. is the standard way of expressing the falsity of any of

them.

The data presented so far is, in my opinion, a powerful reason to abandon the

claim that syntax alone can hold the answer to the mysteries of negation. In the next

45

sub-section, I will be reviewing one important observation from the domain of

semantics.

4.2 Negation as the Binding of Free Variables

From a semantic point of view, negation has been analyzed by various

linguists as an operation of binding free variables:

“I therefore assume, without any further discussion, that standard

sentential negation is binding of event variables by existential closure that is

introduced by the negative operator (following Acquaviva 1997, Giannakidou

1999). Negation, being the negative operator, introduces an existential

quantifier that binds all free variables that have remained unbound during the

derivation.” (Zeijlstra 2004: 179)

In this case, negation will have the following form:

(6) [Op¬ e,x […(e)…(x)…]]35

Zeijlstra goes on to explain that the locus of NegP in a sentence can thus vary

according to the variable it binds: if sentential negation binds an event variable, it will

have to be placed directly above vP, dominating it. If it binds a time variable, it will

dominate and be placed above T0. This can account for the distributional difference

between the place of negation with respect to Tense, formally formulated in the Neg

Parameter by Pollock (1989) and Laka (1990).

Following a proposal by Ramchand (2001), Zeijlstra adopts the view that

languages can differ with respect to which of the variables is bound by the negative

operator: there are languages where negation binds the event variable, languages

where negation binds the temporal variable, and languages that have both strategies.

Such an example, offered by Ramchand and repeated by Zeijlstra, is Bengali. In

Bengali, there are two different negative markers, generally occurring in

complementary distribution: na, the default one, which binds event variables, and ni,

35 Zeijlstra (2004: 179), his example (62)

46

which binds time variables. A strong argument in favor of this is the fact that time

NPI adverbs are only licensed by ni, the negative marker binding the temporal

variable.

From this, we have to retain one conclusion, namely that the syntactic

variation that languages display in their expressing of negation can be accounted for

semantically. What is more, we can agree with Zeijlstra that, if the semantic

explanation is satisfactory and well-grounded, syntactic justifications might not be

necessary after all:

“If the ordering of negative elements with respect to other elements in

the sentences can be explained by a semantic analysis, there is no need to

assume a syntactic selection mechanism as well. (…) I do not take the position

of NegP in the clause to be a result of some (parameterised) syntactic selection

mechanism, but to be driven from its semantic properties. (…) This means that

the locus of negation in the clause follows from semantic properties of

negation.” (Zeijlstra 2004: 178-9)

In the next sub-chapter, I will adopt the view that, for languages lacking

Negative Concord, the role played by syntax with respect to negation might not be

just as important.

4.3. Semantic Negation in Non-Negative Concord Languages

The approach I follow here, as much that has been said so far, comes from

Zeijlstra (2004) and is elaborated in chapter 8 of his thesis (243 onwards).

4.3.1. Syntactic vs. Semantic Negation

In a previous section (2.4), I have touched upon Zeijlstra’s (2004) proposal

that the difference between Negative Concord and Double Negation (or, better said,

non-Negative Concord) is in fact a syntactic / semantic distinction. I include his exact

definitions here:

47

(7) 1. Semantic negation: every negative element corresponds 1:1 to a negative

operator.

2. Syntactic negation: negative elements mark the presence of a (c)overt

negative operator. (Zeijlstra 2004: 244)

This classification is based on the central idea that the way of expressing

negation is not fixed by Universal Grammar, but essentially varies cross-

linguistically, to the extent to which some languages may employ a mechanism that is

altogether different from the one used by other languages.

The definition in (7) 2. is a summary of a rather intricate theory of Negative

Concord, that exceeds the limits of this paper and that I will not describe here. Suffice

it to say that Negative Concord is indeed a syntactic operation and that the negative

elements in the languages that display it are not negative in themselves, but become

negative when they are used in a negative context. The negative operator, overt or

phonetically null, is unique, but its presence is marked by the negative elements.

Following Moscati (2006), we could place this negative operator in the

Complementizer layer36 – he suggests that, from there, it dominates all the negative

elements, through a negative chain.

What interests us now, however, is the strategy for conveying negation in

Non-Negative Concord languages. We established that, in their case, negation is

semantic: negative elements are inherently negative. In the next sub-section, we will

see how Minimalist concepts can help us explain this.

4.3.2. Negation as Feature-Checking

Saying that a particular constituent is or is not inherently negative translates, in

Minimalist terms, to saying that the negative feature it carries is interpretable or

uninterpretable. Interpretable features can be dealt with by LF, but uninterpretable

features cannot, so they must be deleted before Spell Out. This deletion takes place by

checking their uninterpretable feature against an interpretable feature of the same

kind, present in the structure.

36 Haegeman (2000) has already proved that, within Rizzi’s fine-grained structure of the CP, emphaticnegative elements like in “With no job would she be happy” are placed in a Focus Phrase.

48

Feature-checking is a syntactic operation that occurs in Negative Concord

languages. The negative elements here are generated with an uninterpretable negative

feature. The feature of the negative operator, on the other hand, is interpretable.

Feature-checking takes place and the negative meaning of the sentence is ensured37.

But in the case of Non-Negative Concord languages, all negative elements

come from the lexicon with an interpretable negative feature that is submitted directly

to LF. Because they lack the syntactic operation of Negative Concord, interpreting

negation is entirely a matter of semantics and, consequently, they do not have any

negative heads or negative projections, just various adverbs and quantifiers that

express negation by themselves.

The differences we have seen so far can be summarized like this:

(8) 1. Negative Concord negation is syntactic [+uNeg] features

2. Non-Negative Concord negation is semantic [+iNeg] features

A possible counter-argument to this classification could come from the fact

that Non-Negative Concord languages do not have a NegP and an N0, while is has

been agreed that English does have a negative head, n’t38. This is explained by

remembering that the non-Negative Concord status of English is limited both

diachronically and dialectologically, as we have seen in 2.5.

In fact, Zeijlstra shows that English is currently undergoing a process of

changing, with the negative head n’t slowly taking the place of the negative adverb

not, with the latter being nowadays restricted to the formal styles but, in speech, only

appearing when it is focused. This is correlated to losing Double Negation: in those

African American varieties that have Negative Concord, not is no longer available as

a marker of sentential negation, which is rendered solely by n’t (Zeijlstra 2004: 278).

Bringing other arguments, he labels English as a pseudo-Double Negation language

that exhibits some Negative Concord-like behaviors that are connected to the fact that

n’t has become a syntactic head.

Leaving these subtle differences aside, we have to retain the semantic status of

negation in Non-Negative Concord languages:

37 How this is done exactly is still subject to debate. Haegeman (1995 and later) formulates the Neg-Criterion and argues that the operator and the negative head have to be placed in a Spec-Head relation(See Cornilescu (2003 : 52-55) for an application of how this works in English). Linguists vary inexplaining this as the result of either Move or Agree.38 Not is a negative adverb in Spec position.

49

“Negation is flexible cross-linguistically (or language-internally) with

respect to its syntactic categorical status. I have argued that only in a subset of

the set of languages negation is realised as a syntactic category, i.e. it triggers

syntactic operations. In other languages, like Dutch and German, syntax is

blind for negation, i.e. negation does not trigger syntactic operations.

Negative lexical items do not contain formal information for the syntactic

procedure, but consist of material that can be interpreted directly at the

interface with the Conceptual-Intentional component of the language faculty.”

(Zeijlstra 2004: 189)

Let us now summarize the conclusions we have adopted so far:

1. Syntax is not always capable of reasonably explaining negation.

2. Negation is, semantically, the binding of a variable.

3. Non-Negative Concord languages have negative elements with a [+ iNeg]

feature that does not enter syntactic operations.

This proposal is very welcome, as it gives us almost complete freedom to

explain the placing of negation in Non-Negative Concord languages. This will enable

us to explain the presence of negation in the Complementizer layer as the binding of

the whole proposition by a complementizer or another element that comes from the

Lexicon with an interpretable negative feature.

4.4. Negation at the Left Periphery

From this moment on, I will be focusing on non-Negative Concord languages

only. If we established that negation is a purely semantic process and that any of the

variables in a sentence can be bound by it, it is logical to assume that the unique

negative element that produces sentential negation could be placed in any of the three

layers of the clause39, including the Left Periphery. In 4.4.1, I review some arguments

in favor of the view that the CP layer can theoretically host negation. In 4.4.2, I show

39 Already mentioned in chapter 1.1 – the lexical layer (VP), the inflectional layer (IP) and thecomplementizer layer (CP).

50

that it does indeed, and I illustrate this with a few more examples from Latin. 4.4.3. is

a tentative proposal of extending the analysis to English as well.

4.4.1. Theoretical arguments

The arguments that I will mention here are very simple, very logical and, some

of them, very old.

The first comes from Jespersen (1917: 56) who states that natural languages

have two tendencies with respect to choosing the locus of negation: a tendency to

place negation on the verb (which, apart from the fact that “it yields a more elegant

expression”, can be explained by the importance of the VP in the clause) and a second

tendency to “attract the negative notion to any word that can easily be made

negative.” And, as I will show, complementizers and other constituents at the Left

Periphery can easily be negated.

Moreover, there is another very important tendency mentioned by Jesperson40,

that of negating the first negatable element in linear order: “Whenever there is

logically a possibility of attracting the negative element to either of two words, there

seems to be a universal tendency to join it to the first” (cited in Cornilescu 1976: 72).

This is reformulated by Cornilescu as: “In a negative sentence containing several

indefinites (or other elements that might incorporate the Neg), only one constituent is

negated: the first one.” (Cornilescu 1976: 72)

This tendency can be justified by reasons of interpretability – as we find in

Huddleston & Pullum (2002):

“…as the position <of negation> gets further from the beginning of the

clause and/or more deeply embedded, the acceptability of the construction

decreases, simply because more and more of the clause is available to be

misinterpreted as a positive before the negator is finally encountered at a later

stage in the processing of the sentence.” (Huddleston & Pullum 2002: 814)

40 This tendency is clearly at work in non-Negative Concord languages, but possibly not in NegativeConcord ones. Jespersen, however, does not make the difference.

51

I will repeat here the examples that they give, with the mention that the b.

example is just odd, liable to be misinterpreted and difficult to understand, not

actually ungrammatical:

(9) a. I am not satisfied with the proposal you have put to me in any way.

b. ?I am satisfied with the proposal you have put to me in no way.

(10) a. As far as I can recall, I have not purchased food at the drive-through

window of a fast-food restaurant on any street in this city.

b. ??As far as I can recall, I have purchased food at the drive-through window

of a fast-food restaurant on no street in this city.

On the other hand, introducing the negative element as early as possible is the

way for the speaker to make sure that his or her interlocutor will correctly interpret the

sentence or clause as negative. In syntactic terms, “as early as possible” means as high

in the syntactic tree as possible, i.e. in the Left Periphery.

The fact that negation can adjoin to the whole clause has been taken into

account before. In the Standard Analysis framework, Cornilescu (1976) proposes that

negation is concatenated to the entire sentence at Deep Structure, then moves to the

place it occupies at Surface Structure. She gives the following DS representation

(Cornilescu 1976: 60):

(11) S 3Neg S

3NP VP

Despite being theoretically outdated, this representation is a great intuition, as

it shows negation at the Left Periphery, where we will demonstrate it can appear

overtly in Latin and possibly English as well41.

41 And where Moscati (2006) wants to demonstrate it is placed at least covertly in any sentence,regardless of the type of negation (Negative Concord or non-Negative Concord) of the respectivelanguage.

52

4.4.2. Empirical evidence: Latin and other languages

As Moscati (2006) shows in his thesis, there are a few languages, coming from

unrelated language families, that have the possibility of placing negation in the

Complementizer. He gathers supporting data from Latin, Irish, Scottish Gaelic,

Basque, Gbe languages42 and Hebrew.43

What is more, some of these languages not only may, but in fact must place

negation in the Complementizer: “A language like Irish, for example, expresses

negative meanings only by means of its complementizer system” (Moscati 2006: 58).

It would be extremely interesting to see whether the availability (and/or

obligatoriness) of this strategy for expressing sentential negation is in any way related

to the existence of semantic, and not syntactic, negation in these languages44.

For the moment, let us confine ourselves to the analysis of Latin negative

complementizers45.

I consider that Latin has at least three separate negative elements that are

negative complementizers: ne, neve/neu and nisi. Apart from these complementizers

proper, I want to show that the negative copulative conjunction neque/nec can also

find its place in the split CP.

We cannot argue that neque/nec is a complementizer. Its role in the sentence is

not to introduce an embedded clause, but a coordinated sentence. As we have seen in

chapter 3, neque/nec can link either two negative clauses, or a positive clause and a

negative one. Furthermore, neque/nec can even introduce a negative independent

clause, when, at the larger level of the discourse, the sentence is felt to be somehow

linked to the previous ones in the text.

Therefore, we can conclude that neque/nec comes from the Lexicon with the

[+ iNeg] and with a [+copulative coordination] feature. As it always determines the

negative polarity of the whole clause, in other words, it gives a negative force to the

clause, I will assume that its place is always in ForceP:

42 A family of languages spoken in the Western part of Africa.43 Even if Moscati does not mention this, I have reasons to believe that Ancient Greek had negativecomplementizers as well. This remains to be studied.44 Based on Moscati’s data, I could not ascertain if the languages he illustrates are NC languages or not.45 My approach differs from Moscati’s: he only takes ne and quin into consideration, and not the othersI describe. I do not consider quin and quominus to be negative complementizers.

53

(12) Promittis ad cenam nec venis. (Plin. Ep. I, 15, 1)

‘You promise to come to dinner and you don’t come.’

(13) ForceP 3

Force’ 3

Force0 IPnec 3

I’ 3

I0 VP 5

venis

In chapter 3, I mentioned that there are rare case where neque/nec is replaced

by et non, for pragmatic reasons. Now I can explain what those reasons exactly are:

because the speaker wants to draw attention to the negation, the negative marker will

be placed in a sentence-initial Focus position, while the marker of coordination will

remain in the Force phrase:

(14) Videmus… et non commovemur. (Cic., Har., XII, 25)

‘We see… and yet we are not impressed’

(15) ForceP 3

Force’ 3

Force0 FocPet 3

non Foc’ 3

Foc0 IP 3

I’ 3

I0 VP 6

commovemur

54

In the case of ne and neve/neu, we can safely say that they are negative

complementizers46. As shown in chapter 3, they always introduce subordinate clauses

that are constructed with the subjunctive mood. Therefore, we can propose that the

features they come from the Lexicon with are, on the one hand, [+ iNeg], on the other,

[+ irrealis]47. Neve/neu also has a [+ copulative coordination] feature. Because of the

[+ irrealis] feature, I want to suggest that both are placed in the Finiteness phrase:

(16) Cohortatus est…/ ne se admodum animo demitterent/ neve perturbarentur

incommode/. (Caes. Gal., VII, 29)

‘Encouraged his soldiers… that they should not be too much depressed in

spirit, nor alarmed at their loss.’ (trans. W.A. McDevitte)

(17) FinP 3 Fin’ 3 Fin0 IP

ne 3 I’

3 I0 VP 6

demitterent

FinP 3 Fin’ 3 Fin0 IP

neve 3 I’

3 I0 VP 6

perturbarentur

There are also case where ne co-occursss with ut, its positive counterpart, in

order to introduce a negative embedded clause. This is seen by grammarians as an

46 The status of ne as a negative complementizer is proved in Barnes, Brown et. al. (online)47 The irrealis feature was suggested by Avram (p.c.).

55

error, since they decompose ne into et + ut. In fact, the reduplication of the

complementizer can be explained, as above, by assuming different positions for the

two conjunctions: I believe ne will keep its place in FinP, while ut will assume a

higher position, possibly carrying its [+ irrealis] feature to the ForP.

(18) An ego Homero, Ennio, reliquis poetis (…)

concederem ut ne omnibus locis eadem contentione uterentur…?

condone-SUBJ 1ST all places same elevation use-SUBJ IMPERF 3RD PL

‘Or should I condone that Homer, Ennius and the rest of the poets (…) did not

make use of the same elevation everywhere…?’ (Cic. Or., 31,9)

(19) ForceP 3

Force’ 3

Force0 FinPut 3

Fin’ 3

Fin0 IPne 3

I’ 3

I0 VP 6

uterentur

An argument for the low position of ne in the CP is the fact that, if there is

Topicalized or Focused material in a negative embedded clause, this material will

have to be placed in-between the two conjunctions, as we will see in (20): ut has to

remain in sentence-initial position in order to mark the Force of the sentence (and its

relation to the matrix, to whom it is subordinated precisely by means of this ut), while

ne has to stay in FinP to give the verb the [+ irrealis] feature. We might even argue

that, in such cases, it is precisely because of the intervening material (marked by an X

in the (21) representation) that the separation of the two functions of the conjunction

takes place, like in (Cic. Verr., 4. 140).

56

(20) Eos hortatus sum ut causae communi salutique ne deessent.

them urge-PERF 1ST SG cause common salvation-and abandon-SUBJ. 3RD PL

‘I urged them not to abandon our common cause and our salvation’

(21) ForceP 3

Force’ 3

Force0 FocPut 3

X Foc’ 3

Foc0 FinP 3

Fin’ 3

Fin0 IPne 3

I’ 3

I0 VP 6

deessent

The third and last Latin negative complementizer is nisi ‘if not, unless’, the

negative counterpart of si ‘if’, which introduces a conditional clause. It will have the

[+ iNeg] and [+ conditional] features and occupy a position within FinP.

(22) Ortum quidem amicitiae videtis, nisi quid ad haec forte vultis. (Cic. Amic. 32)

see-IND 2ND PL unless want-IND 2ND PL

‘You can now see the origin of friendship, unless you want to add something.’

(23) Fin’ 3

Fin0 IPnisi 3

I’ 3

I0 VP 5

vultis

57

4.4.3. Extending the analysis: English

We have seen the distribution and locus of the negative complementizers in

Latin. Now, assuming that all non-Negative Concord languages behave similarly, we

could try to extend this analysis and see whether English has negative

complementizers. I would like to suggest that at least two English conjunctions could

qualify as negative complementizers: ‘lest’ and ‘unless’.

(24) Be careful lest you should fall.

(25) We will never know the truth unless she tells us.

The first argument would be that, commonly they are explained as having the

same meaning with ‘that… not’ and ‘if… not’, respectively. Since negation in non-

Negative Concord languages is semantic, we could claim a negative meaning for the

two complementizers solely on the basis of the fact that they are understood as being

negative.

With respect to their position, just like their exact Latin counterparts, they will

appear in FinP – ‘lest’ because it caries a [+ irrealis] features and always selects the

subjunctive mood, and ‘unless’ because of its [+ conditional] feature.

(26) Fin’ 3

Fin0 IPlest 3

you I’ 3

I0 VPshould 5

fall

(27) Fin’ 3

Fin0 IPunless 3 she I’

3 I0 VP

5tells

58

4.5. Conclusions

In this chapter, I managed to reach a possible explanation with respect to the

status of negative markers placed at the Left Periphery.

In order to do this, I first had to explain why such a position is available to be

occupied by negative element. I started by providing arguments in favor of the view

that syntax alone cannot provide a satisfactory explanation for the position of negation

– therefore, I presented several ambiguities in scope, in section 4.1.

In 4.2 and 4.3, I provided two approaches to defining negation: from a

semantic point of view, I showed that negation can be conceived of as the process of

binding free variables. Syntactically, as I explain in 4.3.2, negation is a feature-

checking operation. Here, I adopted Zeijlstra’s view that the essential distinction

between Negative Concord languages and languages that are non-Negative Concord is

the type of negative feature they have: if a language has negative elements with

uninterpretable negative features, those elements will have to enter syntactic

relations in order to check their features: this happens in Negative Concord languages.

If, on the other hand, the negative elements come from the Lexicon with an

interpretable negative feature, such a syntactic mechanism is no longer necessary,

since interpretable features are directly processed by LF. This idea gives us the

freedom to place the negative marker wherever it needs to appear in the clause.

In 4.4. I demonstrate that the Left Periphery can be a very good place to have

negation. The main theoretical arguments that I brought in 4.4.1 are related to

interpretability: the higher the negative is, the easier the sentence is perceived as

negative. Therefore, in 4.4.2, I proceeded to show that Latin has overt manifestations

of negation placed in the complementizer layer: for this I re-analyzed neque/nec, ne,

neve/neu and nisi, proposing a possible place for each of them in the fine-grained

structure of the CP. Finally, by analogy, I wanted to suggest that English might have

two such negative complementizers, ‘lest’ and ‘unless’, modeled after ne and nisi

respectively.

All things considered, I want to conclude this chapter with a table

summarizing the findings about negative elements placed at the Left Periphery in

Latin and English:

59

Table 7: Negation at the Left Peripherty

[+iNeg] [+ irrealis] [+copulative coordination]

neque / nec + - +

ne + + -

neve / neu + + +

nisi + + -

lest + + -

unless + + -

Final Conclusions

This paper was my attempt to clarify the case of those negative elements found

at the Left Periphery in non-Negative Concord languages.

In order to do that, I took the data gathered in chapter 3, organized it

according to the concepts, definitions and classifications established in chapter 2,

and integrated it in the theoretical framework I adopted in chapter 1. By doing

this, in chapter 4, I was able to account for the existence of such negative

elements and suggest places for them in the detailed structure of the

complementizer layer.

However, this was only the simple part. The more tedious one would be to try

to find satisfactory approaches that explain the same things for Negative Concord

languages, where negation is syntactic in nature and the uninterpretable features of the

negative element have to be checked against the interpretable feature of a negative

operator. I will have to study and weigh Moscati’s (2006) proposal on that, possibly

applying it to Romanian.

On the other hand, the Left Periphery seems to be a fascinating place to study

– identifying constituents that can occupy various positions there in Latin is on my

academic ‘to do’ list, especially that I am under the impression that some of the Latin

discourse-organizing particles might in fact be overt Foc0 or Top0 heads.

Therefore, I can solemnly state that this paper was, for me, nothing else but a

research-opener.

60

BIBLIOGRAPHY

Avram, L. 2009. Contemporary English Language (English Syntax, II). Lectures.Unpublished

Burrow, J.A., Turville-Petre, Th. 1996. A Book of Middle English. Oxford. Blackwell.

Cornilescu, A. 1976. The Transformational Syntax of English. The Complex Sentence.Bucureşti: Tipografia Universităţii din Bucureşti

Cornilescu, A. 1982. English Syntax. Vol. II. Bucureşti: Tipografia Universităţii dinBucureşti

Cornilescu, A. 2003. Complementation in English. A Minimalist Approach. Bucureşti:Editura Universităţii din Bucureşti

Costa, I. 1999. Neflexibile indo-europene. Bucureşti: Roza Vânturilor

Crystal, D. 2004. The Stories of English. London: Penguin, 2004

de Swart, H., Sag, I.A. 2002. Negation and Negative Concord in French. Linguisticsand Philosophy 25: 373-417

Ernout, A., Thomas, F. 1989. Syntaxe Latine. Paris: Éditions Klincksieck

Haegeman, L. 2000. Negative preposing, negative inversion, and the split CP. InHorn, L. and Kato, Y. (eds.), Negation and Polarity, 21–61. Oxford: OxfordUniversity Press.

Jespersen, O. 1917. Negation in English and other Languages. Copenhagen: A.F.Høst.

Juret, A.C. 1926. Système de la syntaxe latine. Strasbourg: Publications de la Facultédes Lettres de l’Université de Strasbourg.

Manoliu-Manea, M. 1977. Elemente de sintaxă comparată romanică. Tipologie şiistorie. Bucureşti: Tipografia Universităţii din Bucureşti.

Mitchell, B., Robinson, F.C. 2006. A Guide to Old English. Oxford/Cambridge:Blackwell.

Moscati, V. 2006. The Scope of Negation. PhD dissertation, Centro Interdisciplinaredi Studi Cognitivi sul Linguaggio

Oudot, A. J. 1964. Syntaxe Latine. Strasbourg: Publications de la Faculté des Lettresde l’Université de Strasbourg.

61

Pullum, G.K., Huddleston, R. Negation. In Huddleston, R., Pullum, G.K. (eds.). 2002.The Cambridge Grammar of the English Language, 785 – 849. Cambridge:Cambridge University Press.

Rizzi, L. The fine structure of the left periphery. In Rizzi, L. 2000. ComparativeSyntax and Language Acquisition. Routlegde.

Sluşanschi, D. 1994. Sintaxa limbii latine. Vol II, Sintaxa frazei. Bucureşti: EdituraUniversităţii din Bucureşti

Smith, J.J. 2009. Old English. A Linguistic Introduction. Oxford: Oxford UniversityPress.

Touratier, C. 1994. Syntaxe Latine. Peeters: Louvain-la-Neuve.

Touratier, C. 2008. La portée de la negation?. In Bortolussi, B, Brachet J.P. et. al.(eds.), De lingua latina - Revue de Linguistique Latine du Centre AlfredErnout 1 (2008), online

van Gelden, E. 2006. A History of the English language. Amsterdam: J. Benjamin.

Zeijlstra, H.H. 2004. Sentential Negation and Negative Concord. PhD dissertation,Universiteit van Amsterdam.