Negation

9
Negation
Geoffrey K. Pullum
Rodney Huddleston
1 Introduction 786
1.1 Tests for clause polarity 786
1.2 An overview of negation types 787
1.3 Scope and focus of negation 790
1.3.1 The concept ‘having scope over’ 790
1.3.2 Relative scope: wide scope negation and narrow scope negation 792
1.3.3 Focus 796
2 Verbal negation 799
2.1 Primary verbal negation 799
2.2 Imperative negation 802
2.3 Secondary verbal negation 803
2.3.1 Formal marking of secondary negation 803
2.3.2 Secondary negation with not following an auxiliary verb 803
3 Non-verbal negation 806
3.1 Not as a marker of non-verbal negation 807
3.2 The synthetic absolute negators 812
3.2.1 Clausal negation 812
3.2.2 Subclausal negation 815
3.3 The approximate negators 815
3.4 Affixal negation in relation to verbal negation 821
4 Polarity-sensitive items 822
4.1 Negatively-oriented polarity-sensitive items (NPIs) 823
4.1.1 NPIs vs negative idioms 824
4.1.2 Variation in the strength of the negative orientation 825
4.2 Positively-oriented polarity-sensitive items (PPIs) 829
4.3 Correspondences between PPIs, NPIs, and negators 831
4.4 Non-affirmative contexts 834
5 Increased specificity of negation (I don’t want to hear about it) 838
6 Multiple negation 843
6.1 Multiple semantic negation within a single clause 844
6.2 Negative concord and pleonastic negation 845
7 Positive and negative polarity in responses and anaphora 847
7.1 Answers to polar questions and comparable responses 847
7.2 Anaphoric so and not 849
785
Downloaded from https:/www.cambridge.org/core. Boston University Theology Library, on 04 Jun 2017 at 01:27:47, subject to the Cambridge Core terms of use,
available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/9781316423530.010
786
1 Introduction
A pair of clauses such as It is raining and It isn’t raining are said to differ in polarity. The
first is a positive clause or a clause with positive polarity, while the second is a negative
clause or a clause with negative polarity. For the most part positive represents the default
polarity, in the sense that positive constructions are structurally and semantically simpler
than negative ones. To a very large extent, therefore, a description of polarity is a matter
of describing the special properties of negatives – and it is for this reason that we have
called this chapter ‘Negation’ rather than ‘Polarity’.
1.1 Tests for clause polarity

Negation is marked by words (not, no, never, etc.) or by affixes (e.g. ·n’t, un·), but very
often the effect of adding a negative word or the suffix ·n’t is to make the whole clause
negative. Hence the distinction drawn above between It is raining and It isn’t raining as
positive and negative clauses. We will therefore begin by surveying four useful diagnostic
tests for determining the syntactic polarity of a clause. They are illustrated in [1]:
[1] negative clause positive clause
i a. He didn’t read it, not even the abstract. b. ∗He read it, not even the abstract.
ii a. He didn’t read it; neither/nor did I. b. Ed read it; so did I.
iii a. Ed didn’t read it, did he? b. Ed read it, didn’t he?
iv a. Not once did Ed read it. b. After lunch Ed read it.
Single underlining marks the clauses whose polarity is indicated in the headings, while
the double underlining in [i–iii] marks the crucial feature of the diagnostic.
Clause continuations with not even

Negative clauses allow a continuation with not even + complement or adjunct, as in
[1ia]. This is comparable to Ed didn’t even read the abstract, but instead of the abstract
being integrated into the structure of the clause, it is added on, as a prosodically detached
supplement. When the clause is negative, the following even is commonly preceded by
not, as here, but not is inadmissible after a positive clause. (For the meaning of even, see
Ch. 6, §7.3.)
Connective adjuncts
In [1ii] the underlined clause is followed by an anaphorically reduced clause introduced
by a connective adjunct. Following a negative clause we find neither or nor, whereas a
§ 1.2 An overview of negation types 787
positive clause is followed by so. Note that switching the connective adjuncts leads to
ungrammaticality: compare ∗Ed didn’t read it; so did I and ∗Ed read it; neither/nor did I.
A contrast of the same kind applies when the connective adjunct is located at the end of
the following clause:
[2] a. Ed didn’t read it, and I didn’t either. b. Ed read it, and I did too.
Again we may contrast Ed didn’t read it and I didn’t too and ∗Ed read it and I did either.
∗
(The connective adjuncts neither, nor, either and so, too are not restricted to occurrence
in reduced clauses, but the reduced construction provides the simplest test for our
purposes.)
Reversed polarity tags

Did he? and didn’t he? in [1iii] are reduced interrogative clauses, known as tags. They
represent the most common type of interrogative tag, being used to seek confirmation
of what has been said in the clause to which they are attached. This type of tag reverses
the polarity of the preceding clause, so we have negative clause + positive tag in [iiia],
positive clause + negative tag in [iiib].
These are not the only type of clause + tag construction: it is possible to have posi-
tive clause + positive tag, as in Ed read it, did he? and some speakers allow negative
clause + negative tag, as in %Ed didn’t read it, didn’t he? But these are clearly different
intonationally and in their pragmatic effect from those in [1iii], as described in Ch. 10,
§5: our diagnostic is based on the most neutral type of confirmation tag.
Subject–auxiliary inversion with prenuclear constituents

The test illustrated in [1iv] involves the form of the clause itself rather than constraints
on what may follow it. Negative clauses in which the negation is marked on a constituent
in prenuclear position have obligatory subject–auxiliary inversion. Compare [iva], for
example, with ∗Not once Ed read it. There is, by contrast, no inversion in the positive [ivb].
Nor do we have inversion in negative After lunch Ed didn’t feel well since the negation is
marked on the verb, not on the PP after lunch.
Obligatory subject–auxiliary inversion is not limited to negative clauses: inversion is
obligatory following connective so in [1iib] and following a prenuclear phrase introduced
by only (Only occasionally did Ed read these reports) – see Ch. 3, §2.1.2. This test thus
needs to be used in combination with the others, but we will see that it nevertheless
proves useful in drawing the distinction between a negative clause and a positive clause
containing subclausal negation. This is one of the distinctions to which we now turn.
1.2 An overview of negation types

The framework in terms of which we shall describe negation involves four major con-
trasts: verbal vs non-verbal, analytic vs synthetic, clausal vs subclausal, and ordinary
vs metalinguistic. We illustrate in examples [3–6], and then discuss each distinction in
turn.
[3] verbal non-verbal
i a. He doesn’t dine out. b. He never dines out.
ii a. I did not see anything at all. b. I saw nothing at all.
788 Chapter 9 Negation
[4] analytic synthetic

i a. The report is not complete. b. The report isn’t complete.
ii a. Not many people liked it. b. Nobody liked it.
[5] clausal subclausal
i a. She didn’t have a large income. b. She had a not inconsiderable income.
ii a. We were friends at no time. b. We were friends in no time.
[6] ordinary metalinguistic
i a. She didn’t have lunch with my old b. She didn’t have lunch with your ‘old
man: he couldn’t make it. man’: she had lunch with your father.
ii a. Max hasn’t got four children: he’s b. Max hasn’t got four children: he’s
got three. got five.
The first two contrasts have to do with the expression of negation, i.e. with matters of
form, while the second two have to do with meaning, i.e. with the interpretation of
negation.
(a) Verbal vs non-verbal negation

In verbal negation the marker of negation is grammatically associated with the verb,
the head of the clause, whereas in non-verbal negation it is associated with a dependent
of the verb: an adjunct in [3ib], object in [3iib]. This distinction is needed to account for
the occurrence of the auxiliary do, which is required in the [a] examples of [3], but not
the [b] ones. Within verbal negation we then distinguish three subcategories:
[7] i a. You didn’t hurt him. b. You aren’t tactless. [primary]
ii a. Don’t hurt him. b. Don’t be tactless. [imperative]
iii a. It’s important not to bend it. b. It’s important not to be seen. [secondary]
In primary verbal negation the negative marker is associated with a primary verb-form:
here do is required if there is no other auxiliary verb. In imperative verbal negation
do is required even if the corresponding positive does contain an auxiliary verb, as in
[iib]. The third category comprises all constructions other than imperatives containing a
secondary verb-form: infinitivals, subjunctives, gerund–participials, and so on. Auxiliary
do does not occur in these constructions. We label this category secondary with the
understanding that this is a shorthand for ‘non-imperative secondary’.
(b) Analytic vs synthetic negation

Analytic negation is marked by words whose sole syntactic function is to mark negation,
i.e. not and also no in the use in which it contrasts with yes, as in answering a question,
for example. Synthetic negation is marked by words which have some other function as
well. Synthetic verbal negation is marked inflectionally, by negative verb-forms. Synthetic
non-verbal negation is marked by elements of three kinds:
[8] i absolute negators no (including compounds nobody, nothing,
etc., and the independent form none),
neither, nor, never
ii approximate negators few, little; barely, hardly, scarcely; rarely, seldom
iii affixal negators un·, in·, non·, ·less, etc.
§ 1.2 An overview of negation types 789
We treat examples like He had no money as cases of synthetic negation because no

combines the function of marking negation with that of determiner in NP structure.
As a determiner, it expresses quantification. These two functions are separated in the
analytic negative He did not have any money.
The distinction between the absolute and approximate negators is illustrated in:
[9] a. None of them supported her. b. Few of them supported her.
In [a] the number of them who supported her is zero, whereas in [b] it merely approx-
imates to zero – it is located towards the bottom of the scale, in the area that contains
zero. Example [b] in fact has a positive implicature – that some of them supported her.
It nevertheless has important features in common with [a] that motivate its analysis as
negative – note, for example, that the reversed polarity tag for both examples is did they?
The affixal negators are prefixes (un·happy) or suffixes (care·less). The main discussion
of these is in Ch. 19, §5.5, but we look briefly in §3.4 below at the relation between verbal
and affixal negation – at the contrast, for example, between They are not common and
They are uncommon.
(c) Clausal vs subclausal negation

This distinction relates to the tests for polarity given in §1.1: as made clear there, these
are tests that differentiate between negative and positive clauses. Clausal negation is
therefore negation that yields a negative clause, whereas subclausal negation does not
make the whole clause negative. The distinction is seen very clearly when we apply the
tests to such a pair of examples as those in [5ii]:
[10] i a. We were friends at no time , b. ∗We were friends in no time,
not even when we were at school. not even within a few days.
ii a. We were friends at no time , b. We were friends in no time ,
and neither were our brothers. and so were our brothers.
iii a. We were friends at no time , b. We were friends in no time ,
were we? weren’t we?
iv a. At no time were we friends. b. ∗In no time were we friends.
The tests show clearly that We were friends at no time has clausal negation while We were
friends in no time has subclausal negation. The former allows a not even continuation,
takes neither as a following connective adjunct, takes a positive tag, and requires inver-
sion when at no time is placed in prenuclear position. Conversely, We were friends in no
time does not permit not even, takes so as connective adjunct, takes a negative tag, and
does not have inversion when in no time is preposed (the form required is In no time we
were friends).
One further difference is that the example with at no time has an equivalent with
verbal negation whereas the one with in no time does not:
[11] a. We weren’t friends at any time. b. ∗We weren’t friends in any time.
The negation in in no time thus relates only to the phrase, which means “in a very short
period of time”: it doesn’t negate the clause as a whole.
Affixal negation is always subclausal. Compare, for example:
[12] i These terms aren’t negotiable, are they? [verbal negation: clausal]
ii These terms are non-negotiable, aren’t they? [affixal negation: subclausal]
Clausal negation is a matter of syntax, while affixal negation is purely morphological.

For the rest we take clausal negation as the default case, and survey in §3.2.2 the main
places where negative markers other than affixes yield subclausal negation.
(d) Ordinary vs metalinguistic negation.

Consider finally the contrast between the examples in [6]. In [ia] (She didn’t have lunch
with my old man: he couldn’t make it) the negative indicates that it is not the case, not
true, that she had lunch with my old man. But that is not how the negative is understood
in (the intended interpretation of) [ib]: She didn’t have lunch with your ‘old man’: she
had lunch with your father. The latter would normally be used in a context where you
had said She had lunch with my old man. In uttering [ib] I am not disputing the truth
of what you said but rejecting the formulation you used: I’m objecting to your referring
to your father as your ‘old man’. Similarly, She doesn’t live in a /kæsl/, she lives in a /kɑsl/
might be used to reject your pronunciation of castle.
An important case of metalinguistic negation is illustrated in [6iib], Max hasn’t got
four children, he’s got five: here it is used to deny an implicature. We argued in Ch. 5, §5.2,
that Max has four children entails that he has no less than four and implicates that he
has no more than four. Max hasn’t got four children: he’s got three ([6iia]) is thus ordinary
negation. It denies the entailment of the positive, i.e. it indicates that Max has got four
children is false. Max hasn’t got four children, he’s got five ([iib]) denies the implicature. It
is not saying that Max has got four children is false, and hence is not ordinary negation.
Again I would typically use the metalinguistic negative in a context where you had just
said that Max has four children: I am correcting what you said not because it is false but
because it doesn’t go far enough. It is metalinguistic in the sense that it is saying that four
was the wrong word to use (among those yielding a true statement).
1.3 Scope and focus of negation

The notions scope of negation and focus of negation will be introduced together in
this section because they are tightly interlinked. The scope of negation is the part of the
meaning that is negated. The focus is that part of the scope that is most prominently or
explicitly negated. We will explain these concepts by reference to the most elementary
type of negation, verbal negation of declaratives.
1.3.1 The concept ‘having scope over’

The scope of a negative is most easily seen by considering the semantic effect of removing
the negative element. We will illustrate by first exhibiting in [13] a case where the negative
element ·n’t has scope over everything in a sentence, and then in [15] adding to it some
material over which the negative element does not have scope, pointing out the semantic
difference that results.
In the following pair, [a] is negative, and [b] is its positive counterpart:
[13] a. Liz didn’t delete the backup file. b. Liz deleted the backup file.
Each of the clause constituents Liz, deleted, and the backup file in [b] makes a con-
tribution to the meaning; we can therefore give the truth conditions of [b] as a list of
§ 1.3.1 The concept ‘having scope over’ 791
statements:
[14] i “A deletion operation took place”
ii “The deletion operation was performed by Liz”
iii “The deletion operation was performed on the backup file”
Each of these has to be true in order for [13b] to be true. And the falsity of any one of
them is sufficient to make [13a] true. That is, the negated sentence [13a] is true if either
there was no act of deletion, or if any deletion operation that occurred was performed
by someone other than Liz, or if any deletion operation that occurred was performed on
something other than the backup file. The status of the propositions in [14] thus changes
as we switch from [a] to [b]: the falsity of any one of them establishes the truth of [a],
but also the falsity of [b]. All three components are therefore said to be inside the scope
of the negation – or, to put it another way, the negative has scope over all three of them.
The negative can therefore be said to have scope over the whole clause, i.e. over
everything in the clause (except itself): we can say that the scope of the negation here is
“Liz deleted the backup file”.
Now compare the following examples, in which an extra clause has been added to
each of the sentences:
[15] a. Liz didn’t delete the backup file and b. Liz deleted the backup file and
Sue wrote the report. Sue wrote the report.
The truth conditions for [15b] consist of those given in [14] together with
[16] “Sue wrote the report”
But this is also a truth condition for [15a]: if Sue didn’t write her report then both
examples in [15] are false. The status of [16] is thus not affected by the negation: it is
outside the scope of the negative. So the scope of the negative is the same in [15a] as in
[13a], namely “Liz deleted the backup file”.
Notice that [15a] has the form of a coordination of clauses, and it is quite generally
the case that a negative in one clause does not have scope over another clause that is
coordinate with it. There is, though, an exception to this: a negative can have scope over
a clause-coordination that involves gapping (Ch. 15, §4.2). Compare the following:
[17] a. Kim wasn’t at work on Monday b. Kim was at work on Monday
or Pat on Tuesday. or Pat on Tuesday.
These involve the following components of meaning:
[18] i “Kim was at work on Monday”
ii “Pat was at work on Tuesday”
Because of the meaning of or the truth of either one of these is sufficient to establish the
truth of [17b]. But for [17a] to be true, both of [18i–ii] must be false. The negation thus
affects the status of both: both fall within its scope. In this construction, therefore, the
negative has scope over the whole coordination.
Elements whose meaning is not truth-conditional

Not all elements in a sentence contribute to its truth conditions. Consider the connective
adjunct however in:
[19] a. Ed noticed a problem; Liz, however, b. Ed noticed a problem; Liz, however,
didn’t delete the backup file. deleted the backup file.
Here [b] can be analysed into three components of meaning:

[20] i “Ed noticed a problem”
ii “Liz deleted the backup file”
iii “There is some relation of apparent contrast between [i] and [ii]”
The nature of the contrast is not made explicit; one plausible possibility is that the
problem concerns the backup file and the speaker considers Ed’s noticing the problem as
a reason for not deleting the backup file (so that the problem could be investigated). Now
[20iii] does not constitute a truth condition for [19b]: you cannot argue that the latter
was false simply because there is in fact no relation of contrast between the components.
And because it does not affect the truth of the sentence it cannot fall within the scope of
the negative in [19a]: the scope of the negative here is “Liz deleted the problem”, just as
it is in [13a] and [15a].
Semantic and syntactic identification of scope

Scope is in the first instance a semantic concept, and we have been identifying the scope of
negation in semantic terms – saying, for example, that the scope of the negative in [19a]
is “Liz deleted the backup file”. Where a relevant component of meaning is expressed by
a separate syntactic constituent, however, we can equally well refer to it in terms of its
form. In a case like [19a], then, we can say that the clause Ed noticed a problem and the
adjunct however are outside the scope of the negative.
1.3.2 Relative scope: wide scope negation and narrow scope negation
Scope is the semantic analogue of syntactic constituent structure, and in many cases
the syntactic structure reflects the scope of negation in an obvious and elementary way.
Compare:
[21] i a. She didn’t say that she knew him. b. She said that she didn’t know him.
ii a. She didn’t promise to help him. b. She promised not to help him.
In [i] the scope of the negative in [a] is “she said that she knew him”, while in [b] it is
“she knew him”. Thus the negative has scope over say in [ia] but not in [ib], and this
correlates with the fact that in [ia] say is located within the subordinate clause functioning
as complement of the negated verb do, whereas in [ib] the negative is located within the
clause functioning as complement to say. Similarly, the negative has scope over promise
in [iia], but not in [iib], and again the negative is in the matrix clause in [iia] but the
subordinate clause in [iib].
The concept of scope applies to numerous other kinds of element besides negation
markers, including verbs. Compare, for example:
[22] a. She tried to stop offending them. b. She stopped trying to offend them.
In [a] stop is syntactically within the complement of try and semantically within its
scope, and conversely in [b]: here try is within the syntactic complement and semantic
scope of stop. In [21iib], then, it is not simply that promise is outside the scope of the nega-
tive: the negative is inside the scope of promise. We are concerned here, then, with relative
scope: we have two scope-bearing elements and the issue is which has scope over the other.
As far as negation is concerned, the clearest cases are those like [21] where relative
scope is reflected in the contrast between matrix and subordinate clauses.1 Less obvious
1
Even here, however, matters are not entirely straightforward. With auxiliary verbs the issue arises as to
whether a following not belongs in the matrix or the subordinate clause (see §2.3.2), and we will also find
§ 1.3.2 Relative scope 793
are cases where relative scope is not marked by clause subordination in this way. We will
consider two such cases here, involving the scope of negation relative to adjuncts in clause
structure and relative to quantifiers; the issue also arises with respect to coordinators
(see Ch. 15, §2.2.2) and modal auxiliaries (Ch. 3, §9.3.1).
(a) Relative scope of negation and adjuncts in clause structure

Consider first the relative scope of negation and an adjunct such as intentionally.
[23] i Liz intentionally deleted the backup file.
ii Liz intentionally didn’t delete the backup file. [adjunct has scope over negative]
iii Liz didn’t intentionally delete the backup file. [negative has scope over adjunct]
The truth conditions for positive [23i] can be given as follows:
[24] i “Liz deleted the backup file”
ii “Liz did what she did intentionally”
Now [23ii] cannot be true by virtue of [24ii] being false: it can only be true if [24i]
is false. Thus [24ii] is a condition for the truth of both [23i] and [ii] and hence out-
side the scope of negation in [23ii]. And since [24ii] is the meaning contributed by
intentionally, we can say that intentionally in [23ii] is outside the scope of negation. In
[23i] what was intentional was Liz’s deleting the backup file, whereas in [23ii] what
was intentional was Liz’s not deleting the backup file, so the negative is inside the
scope of intentionally: it contributes to specifying what it was that was done inten-
tionally.
The interpretation of [23iii] is quite different. This can be true by virtue of it being
false that Liz acted intentionally. So this time intentionally is inside the scope of negation,
rather than the other way round. An equivalent way of expressing the difference is to say
that in [23ii] the negative has narrow scope relative to the adjunct, whereas in [23iii] it
has wide scope.
We noted that in the most elementary cases relative scope is reflected in a con-
trast between matrix and subordinate clauses, as in [21], and in cases like [23ii–iii] the
difference in meaning can be brought out by means of glosses involving clause subordi-
nation:
[25] i “Liz acted intentionally in not deleting the backup file” [meaning of [23ii]]
ii “Liz didn’t act intentionally in deleting the backup file” [meaning of [23iii]]
When the negative has narrow scope, it appears in the subordinate clause of the gloss;
when it has wide scope, it appears in the matrix clause of the gloss.
Contraries and contradictories
The contrast between [23ii] and [23iii] can also be brought out by noting that they stand
in different semantic relations to the positive [23i]:
[26] i Liz intentionally didn’t delete the backup file and Liz intentionally deleted the backup
file are contraries: they cannot be both true, but they can be both false.
ii Liz didn’t intentionally delete the backup file and Liz intentionally deleted the backup
file are contradictories: they cannot be both true, but they cannot be both false
either.
that in certain circumstances non-verbal negation within a subordinate clause can have scope over the matrix
(see §3.2.1).
Both [23i] and [23ii] are false if Liz deleted the backup file unintentionally, but there is
no context in which both [23i] and [23iii] are false.
Relative scope and the contrast between clausal and subclausal negation
The difference in scope of the negatives in [23ii–iii] is reflected in the fact that while the
latter behaves straightforwardly as a negative clause, the former does not:
[27] narrow scope: subclausal wide scope: clausal
i a. ∗Liz intentionally didn’t delete the b. Liz didn’t intentionally delete the
backup file, and neither did Sue. backup file, and neither did Sue.
ii a. ∗Liz intentionally didn’t delete b. Liz didn’t intentionally delete the
the backup file, did she? backup file, did she?
The neither continuation is permitted in [b] but not [a] (and so . . . would be somewhat
awkward in [a], but the inadmissibility of neither is sufficient to establish a clear differ-
ence). Similarly did she? as a reversed polarity tag seeking confirmation of what is said in
the preceding clause can be added in [b] but not in [a]. Thus [23iii] is a negative clause,
but [23ii] is not.
It should not be assumed, however, that in negative clauses the negative necessarily
has scope over every element in the clause. One case where this is not so is [19a]. Here
the negative does not have scope over however, but that does not prevent the clause being
negative: compare Liz, however, didn’t delete the backup file and neither did Sue (where the
neither continuation shows that the clause is negative), or Liz, however, didn’t delete the
backup file, did she? (where the tag shows the clause is negative). We will see, moreover,
that the exceptions are not confined to cases like this where the element outside the scope
of negation is non-truth-conditional.
Relative scope and linear order
An obvious syntactic difference between [23ii] and [23iii] is that intentionally precedes
didn’t in the former while didn’t precedes intentionally in the latter. The semantic differ-
ence in scope is marked syntactically by a difference in linear order. In both examples,
then, the element with wider scope precedes the one with narrower scope. This represents
the default case:
[28] Given a construction containing two scope-bearing elements, the one which
comes first will generally have scope over the one which comes later.
As implied by the ‘generally’, relative scope does not always correlate directly with
relative order in this way. One factor that may override it is intonation. Compare:
[29] i Liz didn’t delete the backup file intentionally. [negative has scope over adjunct]
ii Liz didn’t delete the backup file – intentionally. [adjunct has scope over negative]
In the intended pronunciations of these, intentionally falls in the same intonational
phrase as didn’t in [i], whereas in [ii] it is prosodically detached. In [i], the default
pattern is observed: the negative has scope over the following adjunct. But in [ii]
the prosodic offsetting of the adjunct allows it to take scope over the whole of what
precedes.
A similar illustration is provided by reason adjuncts:
[30] i Because it cost $50 she didn’t buy it. [adjunct has scope over negative]
ii She didn’t buy it because it cost $50. [ambiguous]
§ 1.3.2 Relative scope 795
In [i] we find the default pattern: the adjunct comes first and has scope over the neg-
ative. The narrow scope of the negative here can again be brought out by a roughly
equivalent sentence in which it appears in a subordinate clause (underlined): The $50
price caused her not to buy it. Here we understand that $50 was too high a price. Example
[ii], however, is ambiguous. It can have an interpretation following the default pattern,
with the negative having scope over the adjunct; here the adjunct will be in the same
intonational phrase as didn’t. In this interpretation we understand that $50 was a good
price, but that this price did not lead her to buy it – and there is an implicature that
she did buy it, for some other reason. But [ii] can also have the same interpretation as
[i]. This departs from the default pattern given in [28], and the adjunct would form a
separate intonational phrase (and in writing it might well be preceded by a comma to
remove the ambiguity).
(b) Relative scope of negation and quantifiers

The issue of relative scope also arises when negation combines with quantification:
[31] i He hasn’t got many friends. [negative has scope over quantifier]
ii Many people didn’t attend the meetings. [quantifier has scope over negative]
Here [i] is the contradictory of He has many friends: one will be true, the other false. But
[ii] is not even the contrary of Many people attended the meetings: both could easily be
true. Given a large enough set of people, it is perfectly possible for the subset attending
and the subset not attending to both qualify as ‘many’. Again the narrow scope of the
negative in [ii] is brought out by a gloss with a negative in a subordinate clause: “There
were many people who didn’t attend the meetings”. Note, however, that [ii] behaves as a
negative clause: cf. Many people didn’t attend the meetings, not even the first one or Many
people didn’t attend the meetings, did they? This confirms the point made above that in
negative clauses the negative does not necessarily have scope over all other elements in
the clause.
Linear order
The examples in [31] follow the default pattern where relative scope matches relative
linear order. Again, however, the default may be overridden under certain circumstances:
[32] i I didn’t agree with many of the points he made. [scope ambiguous]
ii Everybody didn’t support the proposal, but most did. [wide scope negation]
Example [i] can be interpreted with either the negative or the quantifier having scope
over the other. With wide scope negation it is like [31i], the contradictory of I agreed with
many of the points he made. In this reading there weren’t many points that I agreed with.
The wide scope quantification reading can be expressed unambiguously by fronting the
quantified NP: Many of the points he made I didn’t agree with. In this second reading
there were many points that I disagreed with; it is a less likely reading of [32i] than
the one with wide scope negation, but certainly possible, especially with a clear change
of intonation contour on many. Without the but clause, [32ii] would be ambiguous in
the same way: most did, however, forces the interpretation where the negative has scope
over the quantifier, as it unambiguously does in Not everybody supported the proposal.
Again the intonation can assist in making the meaning clear: the reading with wide scope
negation reading will typically be encouraged by high pitch on EVerybody.
Because the preferred (most likely) interpretation matches scope with order of ap-
pearance in the sentence, if we change the order – for example, by switching from active
to passive – we will change the preferred interpretation:
[33] i Many members didn’t back the proposal. [narrow scope negative]
ii The proposal wasn’t backed by many members. [wide scope negative preferred]
A prosodic override is virtually impossible in [i]: many has scope over the negative. In
[ii] the preferred reading has many within the scope of the negative (“there weren’t many
who supported the proposal”), though it is just possible for this to be overridden, making
it equivalent to [i].
Equivalence between wide scope universal and narrow scope existential quantification
[34] i All of them didn’t have a clue what he meant. [wide scope universal]
ii None of them had a clue what he meant. [narrow scope existential]
These are semantically equivalent (cf. Ch. 5, §5.1). In [i] the universal quantifier all has
scope over the negative: all of them had a negative property. In [ii] none expresses the
negation of existential quantification: “it is not the case that any of them had a clue what
he meant”. (A more emphatic version is Not one of them had a clue what he meant, with
negative and quantifier expressed separately, and the one with wider scope coming first.)
Although [34i–ii] are equivalent, version [ii] is quite strongly preferred. This preference
for a formulation with existential quantification within the scope of the negative over
universal quantification with scope over the negative is reflected in the possibilities for
overriding order in clauses combining negation with universal quantification:
[35] i All of the members didn’t support the proposal.
ii The proposal wasn’t supported by all of the members.
Here [i] can be interpreted with wide scope negation (“Not all of the members supported
the proposal”): it allows a prosodic override of the narrow scope negation reading much
more readily than [33i] because None of the members supported the proposal would be
preferred over the narrow scope negation reading of [35i]. Conversely, override is hardly
possible in [ii]. The normal reading here has wide scope negation (“not all”): instead of
overriding the order to put all outside the scope of negation one would normally use any
instead of all (The proposal wasn’t supported by any of the members), with the negative
having scope over an existential quantifier.
1.3.3 Focus
In all but the most trivial negative clauses there are several different conditions whose
failure to hold would cause the clause to be strictly true. Which condition is intended
can be indicated by a speaker through the device of stressing the most closely associated
word. A constituent marked by stress as being crucial to the way in which an instance of
negation should be understood is called the focus of that negation.2
Negation focus, falsity conditions, and prosody

Let us again compare a negative clause and its positive counterpart:
[36] a. Your children don’t hate school. b. Your children hate school.
2
Throughout this chapter ‘focus’ is to be understood in the sense of ‘informational focus’: see Ch. 6, §7.3, for
the distinction between two concepts of focus.
§ 1.3.3 Focus 797
One way of giving the truth conditions for the positive [b], one which highlights the
contribution of the separate words, is as follows:
[37] i a. “Somebody’s children hate school” b. “You are that person”
ii a. “Some relatives of yours hate school” b. “They are your children”
iii a. “There is some attitude your children b. “That attitude is hatred”
harbour towards school”
iv a. “There is something your children hate” b. “That thing is school”
In order for [36b] to be true, each of these conditions must be true; but for [36a] to be true
it is sufficient that any one of the conditions be false. From this point of view, negative
statements run the risk of being relatively uninformative. It could be that somebody’s
children hate school, but not yours – or indeed that no one’s children hate school; it
could be that some relatives of yours hate school, but not your children – or indeed that
none of your relatives hate school. And so on.
English provides ways of making negatives more informative by giving some indica-
tion as to which condition fails to hold. The way we are concerned with here involves the
use of stress and intonation to highlight the part of the clause that is associated with that
condition. Consider, for example, the four ways of saying [36a] shown in [38], where
the small capitals indicate heavy stress and raised or changing pitch:
[38] i YOUR children don’t hate school.
“If there are children who hate school, they are not yours”
ii Your CHILDREN don’t hate school.
“If any of your relatives hate school, it is not your children”
iii Your children don’t HATE school.
“If your children harbour an attitude towards school, it is not hatred”
iv Your children don’t hate SCHOOL.
“If your children do hate something, it’s not school”
The part of the clause that is prosodically highlighted is the focus. In [38i] the focus is
your, and this indicates that what makes the positive Your children hate school false is the
non-satisfaction of the condition associated with your. Hence the gloss we have provided
for this reading, “If there are children who hate school, they are not yours”, which locates
the failure to satisfy the set of conditions given in [37] specifically in [ib]. Similarly, the
glosses given for [38ii–iv] reflect the choice of children, hate, and school respectively as
focus.
Narrower and broader negation focus

In this very simple example we have taken the focus to be simply the stressed word itself.
But it does not have to be just that word: it can be a constituent that includes the stressed
word. The focus in Your CHILDREN don’t hate school, for example, could be the whole NP
your children, and the interpretation in this case would be “If there are some people who
hate school, they are not your children”. Here, then, the focus is broader than we took
it to be in our interpretation of [38ii]. Similarly, the focus in Your children don’t hate
SCHOOL need not be as narrow as school: it can also be the VP hate school, and here the
interpretation would be along the lines of “If your children have some property, it’s not
that of hating school”.
To illustrate this variation in how broad a focus is selected, consider the following
more complex example:
[39] i At least Max didn’t wear a green mohair SUIT to the wedding.
ii a. “At least the green mohair garment Max wore to the wedding wasn’t a suit”
b. “At least the green garment Max wore to the wedding wasn’t a mohair suit”
c. “At least the garment Max wore to the wedding wasn’t a green mohair suit”
d. “At least what Max did wasn’t to wear a green mohair suit to the wedding”
e. “At least Max wearing a green mohair suit to the wedding didn’t happen”
We confine our attention to the case where the stress is placed on suit. This allows
five different choices of focus, each a constituent containing the word suit. The glosses
given in [ii] correspond to progressively broader foci. In [iia] the focus is taken to be
suit; [iib] makes mohair suit the focus; [iic] makes it green mohair suit; [iid] makes it
wear a green mohair suit to the wedding; and [iie] makes it the whole clause. The later
paraphrases are probably more plausible in normal contexts than the first two, but all
are possible.
There are default assumptions for both main stress and focus of negation. The neutral
place in a clause to put the heaviest stress is on the last stressed syllable of the lexical
head of the last phrasal constituent of the VP. When this is the stress pattern, the focus
is quite likely to be taken by the hearer to be the whole clause that corresponds seman-
tically to the scope of the negation. Consider a normal pronunciation of the sentence I
don’t know why they appointed him to the job. The heaviest stress is likely to be on job,
and given that stress the focus of the negation will probably be taken to be the entire
scope of not, i.e., the whole clause meaning “I know why they appointed him to the
job”.
Positive implicatures resulting from the choice of focus

The focus of negation serves to narrow down the condition whose non-satisfaction makes
the negative true and the positive counterpart false. But in picking out one condition
one often implicates that the other conditions for the truth of the positive are in fact
satisfied. The glosses we gave in [38] did not incorporate these implicatures. We glossed
YOUR children don’t hate school as “If there are children who hate school, they are not
yours”. But very often it will be interpreted more strongly as “There are children who
hate school, but they are not yours”. Thus in addition to the negative component “they
are not yours”, we have the positive implicature “there are children who hate school”. It
is clear that this is an implicature, not an entailment: to select your as focus is not to
actually say that somebody else’s children hate school. It would be perfectly coherent,
for example, to say I don’t know about other people, but I’ll certainly concede that YOUR
children don’t hate school.
Such positive implicatures tend to be stronger with narrow focus than with broad.
Thus if I haven’t done my TAX return yet, say, is interpreted with the whole clause as focus
it will not convey any positive implicatures of this kind. Compare this with a narrow
focus example discussed earlier, She didn’t buy it because it cost $50 ([30ii]). The reading
where the negative has scope over the reason adjunct will normally have the stress on
$50, and focus on because it cost $50. It will then have a very strong positive implicature
that she did buy it, giving the interpretation that she bought it for some reason other
§2 Verbal negation 799
than its price. (But again, this is not an entailment: it could be that I don’t know whether
she bought it or not but am merely quite confident that she would not have allowed the
mere price to induce her to buy it.)
Focus of negation as a special case of informational focus

The concept of focus applies in all main clauses, not just negatives. Compare:
[40] i Liz INTENTIONALLY deleted the backup file.
ii Liz INTENTIONALLY didn’t delete the backup file.
iii Liz didn’t INTENTIONALLY delete the backup file. [focus of negation]
In all three cases the focus is intentionally, and in all three cases the effect of selecting
this as focus is to give prominence to the associated piece of information – that Liz did
what she did intentionally. Example [i] is positive and in [ii] the scope of the negation,
“deleted the backup file” does not include the focus. Only in [iii] does the focus fall
within the scope of a negative, but when it does, it serves as the focus of the negation.
The effect of selecting as focus a constituent within the scope of negation is still to give
prominence to the associated piece of information, but by virtue of giving it prominence
I indicate that this is where you are to find the condition whose non-satisfaction makes
the negative true.
2 Verbal negation
Negation of a clause is commonly marked on or adjacent to the verb of that clause, and
we call that verbal negation. There are three types of clause that exemplify it, and the
syntax is different in each case. We will deal first with clauses that have verbs inflected in
a primary inflectional form, then with imperative clauses, and then with non-imperative
clauses whose verbs are inflected in secondary forms.
2.1 Primary verbal negation

Positive clauses containing a primary form of an auxiliary verb may be negated by
adding not after the verb, giving analytic primary negation, or by inflecting the verb in
the negative, giving synthetic primary negation:
[1] i Kim will be here later on. [positive clause with auxiliary verb]
ii Kim will not be here later on. [analytic primary negation]
iii Kim won’t be here later on. [synthetic primary negation]
In [i] we have the neutral present tense form will in a positive clause. In [ii] this neutral
form will is modified by not in post-auxiliary position. In [iii] the negation is marked
by the negative present tense form won’t.
To negate a clause containing a primary form of a lexical verb it is necessary to add
the semantically empty auxiliary that we have called supportive do:
[2] i Kim waved to us. [positive clause with lexical verb]
ii Kim did not wave to us. [analytic primary negation]
iii Kim didn’t wave to us. [synthetic primary negation]
Do takes on the inflectional properties (here preterite tense) that in the positive are
carried by the lexical verb, and the latter now appears in the plain form. Again the
negation may be marked analytically or synthetically by a negative form of do.3
The choice between analytic and synthetic primary negation

Analytic and synthetic primary negation forms are not fully interchangeable. The main
difference is one of style level.
The synthetic forms are a mark of informal style. While they are the default form
in ordinary conversation and informal writing,4 they are not used in very formal and
solemn contexts or in some kinds of written (especially published) language. They are by
no means absent from academic prose, but the author or speaker who uses them makes
a definite style decision: the effect of using synthetic negative auxiliaries is to increase
the sense of familiarity, intimacy, and accessibility. Compare the following pairs, where
the [a] members belong to formal style, the [b] members to informal style:
[3] i a. I do not accept, and will not condone or defend, this shameful policy.
b. I don’t accept, and won’t condone or defend, this shameful policy.
ii a. This is not to say that one could not conceive of a world in which aesthetic
properties did not supervene on the physical; but the necessity of positing them
does not seem to me an attractive prospect.
b. This doesn’t mean that you couldn’t imagine a world where aesthetic properties
didn’t supervene on physical ones; but having to assume them doesn’t seem like
an attractive prospect to me.
While [ia] sounds grave and parliamentary, [ib] does not, and might be judged to
sound too petulant in the context of a political speech. Likewise, [iia] has the tone of a
published paper in analytic philosophy, while [iib] (in which various other changes are
made in addition to the choice of synthetic negatives) has the flavour of a more informal
explanation of the same ideas; it might be judged too patronising for appearance in
some philosophy journals, but just right for a lecture. Naturally, there is much variation
in style, and some academic writers choose to use a much more informal style than
others.
The fact that synthetic negative auxiliaries are informal does not mean that analytic
forms are neutral or preferable. Analytic forms sound unnatural in many conversational
contexts unless there is some clear reason for their use, e.g., emphasis on the word not
(I did NOT sneak out by the back door when she arrived!). Thus in an ordinary conversation,
I don’t think so or Don’t worry, I won’t be long are perfectly natural even in slow and careful
speech, while I do not think so or Do not worry, I will not be long would be not just unusual
but highly unnatural.5
3
In earlier forms of English, non-auxiliary verbs in primary forms were also postmodified by not. Instances
are preserved in various biblical and proverbial phrases (I care not whether she lives or dies ; He who knows
not, and knows that he knows not, can be taught), but the construction is no longer part of the productive
syntax of the language.
4
This applies to the synthetic forms that are fully acceptable; for the case of %mayn’t and of aren’t with a 1st
person singular subject, see Ch. 18, §5.5.
5
Such unnatural avoidance of negative auxiliaries is used in films to underscore the alienness of such characters
as visitors from outer space.
§ 2.1 Primary verbal negation 801
Grammatical restriction on synthetic negatives: inadmissible in inverted conditionals

Although the difference is mainly stylistic, there is one construction that does not permit
synthetic negatives, namely the inverted conditional (see Ch. 3, §2.1.2). Compare:
[4] i Had it not been for the weather, the plan would have succeeded.
ii ∗Hadn’t it been for the weather, the plan would have succeeded.
Synthetic verbal negation is a matter of inflection
The suffix ·n’t is often mistaken for a reduced pronunciation of the word not. That is
of course its etymology, but it is no longer an unstressed pronunciation of not; in the
contemporary language it is a verbal suffix. There are syntactic, morphological, and
phonological reasons for distinguishing between such pairs as does not and doesn’t, as
described in Ch. 3, §1.9.
Position of negator in clauses with subject–auxiliary inversion

Clauses with subject–auxiliary inversion normally have the subject immediately follow-
ing the auxiliary verb. The negative marker will thus precede the subject in synthetic
negation but follow the subject in analytic negation:
[5] i a. She doesn’t agree with me. b. She does not agree with me.
ii a. Doesn’t she agree with me? b. ∗Does not she agree with me?
iii a. ∗Does shen’t agree with me? b. Does she not agree with me?
Example [iiia] is ungrammatical for the reasons just given: ·n’t is an inflectional suffix of
the verb and hence can never be separated from the verb. In general the word not cannot
come between the auxiliary and the subject, as shown in [iib]. The construction with
not preceding the subject is not competely excluded, however: it is occasionally found as
an alternant of the normal pattern where not follows the subject. Compare:
[6] i Do most self-indulgent public officials not accept bribes?
ii Do not most self-indulgent public officials accept bribes?
Construction [ii] is a survival of an older pattern where not was quite generally permitted
in this position. In speech it would be highly unnatural except in extremely formal
declamation. But in writing it will still be found in sources not permitting synthetically
negated auxiliaries. It is normally restricted to cases where the subject is relatively long,
where it serves to avoid the lengthy interruption between auxiliary verb and not which
makes [i] too sound somewhat stilted.6
Subclausal primary negation

Primary negation is normally clausal. Subclausal cases are restricted to constructions
where the negated verb falls within the scope of a preceding adjunct, as in He often isn’t
there when you call him. The reversed polarity tag would be isn’t he?, and we could have
continuations like and so is his secretary. What is negated here, then, is the predicate (that
is, the VP), not the whole clause.
6
We have noted, however, that synthetic negatives are not permitted in inverted conditionals, and this may
facilitate the use of the construction with not preceding the subject: She might have regretted her smallness
had not all the parts been so well-proportioned.
2.2 Imperative negation

Imperative clauses have a pattern of negation that cannot be reduced to the gen-
eralisations about the other types of negation, uniformly using the negative auxil-
iary don’t. Positive and negative examples of the main types of imperative are shown
in [7]:
[7] positive imperatives negative imperatives
i a. Look at me. b. Don’t look at me.
ii a. You look at me! b. Don’t you look at me!
iii a. Everyone shout it out. b. Don’t everyone shout it out.
Contrasts between imperatives and non-imperatives

Verbal negation in imperatives differs from that in other clauses in the following
respects:
(a) Auxiliary do required even with auxiliaries
[8] imperative declarative
i a. Don’t be afraid. b. You aren’t afraid.
ii a. Don’t have eaten all the pizza by b. I hope [they haven’t eaten all the pizza
the time I get back. by the time I get back].
Here auxiliary do appears in imperatives but not in comparable declaratives – because
they contain auxiliary verbs, which permit primary verbal negation without the addition
of supportive do. The most usual case of this kind involves the verb be, as in [i]. In [ii]
we have the perfect auxiliary have; this is quite rare in imperatives, so [iia] may seem
somewhat contrived. Nevertheless, its grammatical status is sharply different from that
of the clearly ungrammatical ∗Haven’t eaten all the pizza by the time I get back! The
distinction between auxiliary verbs and lexical verbs is thus irrelevant to the formation
of negative imperatives: do is required in all cases.7
(b) Order of subject and verb

The subject, in imperatives that have one, as in [7ii–iii], precedes the verb in the positive,
but usually follows don’t in the negative (but see also Ch. 10, §9.7.3). No such order
change accompanies negation in other clause types.
(c) Synthetic negation is found even though imperatives have plain form verbs
The verb in imperative clauses is in the plain form: it is not a tensed form, as evident
from the form of be in Be careful and the absence of the ·s suffix on shout in [7iii]. The
negative imperative is the only construction containing an inflectional negative that is
not a primary verb-form.
(d) Restrictions on analytic negation

In speech, verbal negation in imperatives is generally expressed synthetically, with don’t.
Analytic do not occurs mainly in writing and is a somewhat stronger marker of for-
mal style than analytic negation is with primary verb-forms. It is particularly unlikely
when the imperative has an overt subject, and in the case of you it can be regarded
7
Verbally negated imperatives without do will occasionally be encountered, but they are conscious archaisms,
surviving as relics, usually biblical, literary, or proverbial: Fear not; Be not afraid; Waste not, want not; Judge
not that ye be not judged; etc. This construction is comparable to that illustrated in footnote 3.
§ 2.3 Secondary verbal negation 803
as ungrammatical:
[9] i a. Don’t any of you think you have b. Do not any of you think you have
heard the last of this matter. heard the last of this matter.
ii a. Don’t you renege on our deal. b. ∗Do not you renege on our deal.
Note that there is no alternative version of the analytic construction where not follows
the subject. This order is found only in interrogatives: Do you not habitually renege on
your promises?
2.3 Secondary verbal negation

The final category of verbal negation covers clauses other than imperatives that contain
a secondary verb-form: a plain form, past participle, or gerund-participle. Since imper-
atives have plain form verbs, this category might be called non-imperative secondary
negation, but we will simplify by dropping the ‘non-imperative’ and speak of secondary
negation. The category covers subjunctive and non-finite clauses with verbal negation.
2.3.1 Formal marking of secondary negation

Secondary verbal negation differs from the others in two respects. First, it never intro-
duces the auxiliary verb do. Second, it is always analytic: negative verb-forms are all
primary except for the don’t of imperatives. In all cases, then, secondary verbal negatives
are formed by placing not as premodifier of the VP, as in the following subordinate
clauses (enclosed in square brackets):
[10] positive negative
i a. It is vital [that he be told]. b. It is vital [that he not be told].
ii a. [Locking the doors] is unwise. b. [Not locking the doors] is unwise.
iii a. [His accepting it] was a shock. b. [His not accepting it] was a shock.
iv a. a plan [approved by the board ] b. a plan [not approved by the board ]
v a. It looks bad [for them to smile]. b. It looks bad [for them not to smile].
vi a. They let me [wear high heels]. b. They let me [not wear high heels].
Here [i] is subjunctive, [ii–iii] gerund-participial, [iv] past-participial, [v–vi] infinitival.
The infinitival with to has an alternant where not follows to: It looks bad [for them to
not smile]. Since to carries no identifiable contribution to meaning, there is no real
possibility of any scope contrast between the two constructions: they are semantically
equivalent. The one with not between to and the verb is a special case of the ‘split infinitive’
construction discussed in Ch. 6, §7.1.
Most clauses with secondary negation are subordinate clauses, but the minor clause
types that have secondary verb-forms in main clauses with exclamatory or optative senses
also show this pattern:
[11] i a. A letter written on a computer! b. A letter not written on a computer!
ii a. My only son getting into Harvard! b. My only son not getting into Harvard!
iii a. Oh to have to visit England! b. Oh to not have to visit England!
2.3.2 Secondary negation with not following an auxiliary verb

When not comes between two verbs the issue arises as to which of them it negates. If the
first verb is a lexical verb or a secondary form of an auxiliary, not belongs clearly with
the following verb:

[12] i She agreed [not to make a formal complaint].
ii Jill’s instruction had been [not to take on any extra staff ].
These are straightforward cases of secondary negation, with not located syntactically in
the infinitival clause. The matrix clauses with primary verb-forms are positive, as evident
from the reversed polarity tags didn’t she? and hadn’t it?, and other tests. Not can modify
a preceding verb only if it is a primary form of an auxiliary (or imperative do). This is
the case we need to consider further.
(a) Not immediately following a primary form of an auxiliary

In this position not normally modifies the auxiliary verb and hence marks primary
negation. This is so even when the auxiliary is not within the semantic scope of the
negative:
[13] a. They must not read it. b. They need not read it.
In [a] must has scope over not (“it is required that they not read it”), whereas in [b] not has
scope over need (“it is not necessary for them to read it”) – cf. Ch. 3, §9.3.1. Nevertheless,
not belongs syntactically in the matrix clause in both cases. Note, for example, that they
take positive tags (must they?, need they?) – and that the analytic negation alternates with
synthetic mustn’t and needn’t.
There are, however, three exceptions to the general rule, illustrated in:
[14] i Jill’s instruction was [not to take on any extra staff – in any circumstances].
ii You can [not answer their letters]: you’re not legally required to respond.
iii You can’t [not go with them].
Not after be in its specifying sense
Example [14i] can be interpreted like [12ii] without the perfect tense feature: “Jill’s
instruction was that they not take on extra staff ”. The reversed polarity tag in this
case would be wasn’t it?, for the main clause is positive. The construction illustrates
the specifying use of be with a clause in the complement position that happens to be
subjectless and negative. In [14i] it is impossible to replace was not by wasn’t and preserve
the sense.
However, it is perfectly possible for the not to follow be in sentences of superficially
similar form and belong to the copular clause (if the copula is in a primary form):
compare the attested sentence The sole purpose of the criminal law is not to amuse
Mr Mortimer, which we understand as denying that the purpose of the criminal law
is to amuse Mr Mortimer, not as affirming the bizarre claim that its purpose is to avoid
amusing him. The construction is therefore potentially ambiguous. We included the
prosodically detached in any circumstances in [14i] because it strongly favours the sec-
ondary negation reading that we wish to illustrate.
Can not and possibility of abstention
The meaning of the first clause in [14ii] is “You are permitted not to answer their letters”,
with can thus having scope over not. In this use, the not will characteristically be stressed
and prosodically associated with answer rather than can by means of a very slight break
separating it from the unstressed can. The fact that the modal has scope over not makes
this semantically comparable to [13a], but syntactically it differs in that it has secondary
negation rather than the primary negation of [13a]. This is evident from the fact that
§ 2.3.2 Secondary negation with not and auxiliary verb 805
the reversed polarity tag for the can clause is can’t you? The meaning is quite different
from that of You can’t/cannot answer their letters, which has primary negation in the can
clause, and where the negative has scope over the modal (“It is not possible or permitted
for you to answer their letters”).
In [14ii], therefore – unlike [13a] – the syntax matches the semantics, with the not
located in the subordinate, non-finite, clause. But this construction is fairly rare, and
sounds somewhat contrived. It would not normally occur with modals other than can
and may.
Can’t not: secondary negation combined with primary negation
In [14iii] (You can’t not go with them) the not follows a negative auxiliary: a verb cannot
be negated twice, so the not can only belong in the non-finite clause. Note that the matrix
clause is straightforwardly negative by virtue of the synthetic negation: the reversed
polarity tag is can you?, just as it is for You can’t go with them. Syntactically, two negatives
do not cancel each other out to make a positive: there are two negative elements in [14iii]
but the main clause remains syntactically negative.
In principle, this construction is possible with any auxiliary verb: I won’t not speak
this time, I promise you. It can even be used, and has been attested, when the auxiliary
is supportive do, to deny a negative assertion that has been made: I DIDN ’T not listen to
you. Compare also imperative Don’t not go just because of me. These examples have one
negative immediately within the scope of another, so that they cancel each other out
semantically. I won’t not speak is truth-conditionally (but not rhetorically) equivalent
to I will speak; I DIDN ’T not listen is similarly equivalent to I DID listen; Don’t not go is
equivalent to Go. In most circumstances the simpler positive would be preferred, but
special circumstances can make the use of mutually cancelling negatives preferable.8
The can case in [14iii] differs from these in that the first negative has scope over can, so
that You can’t not go with them is not equivalent to You can go with them; it is equivalent
to You must go with them, but there is sufficient pragmatic difference between can’t not
and must to motivate the use of both forms.
(b) Not syntactically separated from an auxiliary

Not can more readily mark secondary negation if it is adjacent to the secondary verb-
form but not to the auxiliary. There are three distinct ways in which such a separation
of the auxiliary from the negator can come about.
Infinitival to preceding not
The subordinator to can separate an auxiliary from not. One very clear case concerns
infinitival to in the specifying be construction. Compare:
[15] a. Their aim is not to change things. b Their aim is to not change things.
8
Three attested examples of mutually cancelling adjacent negative elements of this kind are found in the
following interesting piece of dialogue from a film script. The situation is between a man, A, who has run
into a woman, B, whom he had been dating but has not seen for a while.
A: Don’t think that I have [not called you]. I haven’t [not called you]. I mean . . . , I don’t mean that I haven’t
[not called you] because that’s a double negative so as to say that I have called you. . .
B: When did you call?
A: I didn’t. But I didn’t [not call you] in the way that you might think that I didn’t call you. [i.e. as indicating
that I didn’t want to see you again]
Each occurrence of not is prosodically associated with call, and marks secondary negation in the non-finite
clause functioning as complement of the auxiliary.
In [a] we find the ambiguity between primary negation in the matrix clause and secondary
negation in the infinitival clause that we mentioned in the discussion of [14i]; this
ambiguity is resolved in favour of primary negation in the matrix if is not is replaced by
synthetic isn’t. In [b], by contrast, only the secondary negation reading is possible: the
subordinator to marks the beginning of the infinitival VP of the subordinate clause, so
not can only be in that subordinate clause.
Adjuncts between auxiliary and VP
A second case has an adjunct between the auxiliary and the not:
[16] i You can simply not answer their letters, can’t you?
ii They have always not enforced that regulation, haven’t they?
The intervening adjunct serves to dissociate the not from the auxiliary. Compare [16i]
with [14ii] above: adding the simply facilitates the association of not with the non-finite
clause. In [16ii] the universally quantified adverb always has scope over not, forcing the
association of not with enforce, to give the meaning “waive”.
Subject between auxiliary and not
A third possibility is for the subject to intervene between the auxiliary and its comple-
ment, when the matrix clause has subject–auxiliary inversion:
[17] i Would you not put your feet on the sofa.
ii Can you not ask them to help you?
iii Did you not agree with her?
The natural interpretation of [i] is as a request not to put your feet on the sofa: not
belongs in the put clause, and there is no alternant with synthetic negation in the matrix
clause (i.e. we can’t replace would you not by wouldn’t you). But [ii] is ambiguous. It can
be interpreted like [i], i.e. as a request not to ask them (in which case a full stop might be
preferred to the question mark); this reading would be forced if please were inserted
before not. Or it can be interpreted with not having scope over the modal, equivalent to
synthetic Can’t you ask them to help you?; in this reading we have primary negation in
the matrix clause.
Strictly speaking, [17iii] is likewise ambiguous between primary negation of the matrix
and secondary negation of the complement. The latter reading would be assisted if there
were a slight prosodic separation of you from not, as in [14ii], but this reading is strongly
disfavoured as the relevant meaning could be expressed much more clearly by Did you
disagree with her? Again we could insert an always to force the secondary negation
construal: Did you always not agree with her?
3 Non-verbal negation
We turn now to constructions where the negator is not associated with the verb. We look
first in §3.1 at analytic negation marked by not. Then in §§3.2–3 we consider synthetic
negation marked by absolute and approximate negators respectively. Finally, in §3.4 we
discuss the semantic difference between affixal negation and verbal negation. Analytic
negation marked by no (as in answers to questions) is discussed in §7.
§ 3.1 Not as a marker of non-verbal negation 807
3.1 Not as a marker of non-verbal negation

The clearest cases where not marks non-verbal negation are illustrated in:
[1] i [Not all of them] regarded it as a success.
ii He seemed [not entirely honest].
These clauses contain primary forms of lexical verbs, which are not permitted with verbal
negation: compare the verbal negation constructions They did not all regard it as a success
and He did not seem entirely honest (which have synthetic alternants containing didn’t).
When we turn to constructions with auxiliary verbs a contrast emerges because of
the very strong tendency for not after an auxiliary to be interpreted as primary negation:
[2] i [Not all of them] had regarded it as a success.
ii He was not entirely honest.
The negation in [i] remains clearly non-verbal: not is part of the subject NP, and not in
a position where there could be any doubt about that. In [ii], however, not follows the
auxiliary, and the sentence is most likely to be interpreted as a case of primary verbal
negation. Note that the reversed polarity tag for [1ii] would be didn’t he?, whereas that
for [2ii] would be was he?
As in the case of [17] in the previous section, our syntactic description does in fact
entail that [2ii] is structurally ambiguous, with not marking either verbal or non-verbal
negation, because we have said nothing to exclude the bracketed AdjP constituent in [1ii]
from occurring as the complement of the copula. And indeed, the non-verbal negation
reading can be forced if the copula is separated from the not; thus we might find He
was both not entirely honest and somewhat aggressive.9 Nonetheless, the likelihood that
an example like [2ii] will be seen as ambiguous is very low in most contexts.
We now proceed to review a dozen constructions where things are not as subtle as
this, and not can be shown to be quite clearly a marker of non-verbal negation. (An
additional one where not functions as an anaphoric complement, as in I think not, is
covered in §7.2.)
(a) The not all type

[3] i Not all people have had the opportunities you have had.
ii Not often do we see her lose her cool like that.
∗
iii I agree with not all your arguments.
∗
iv He not often visits his parents.
Not combines with a quantifier to form a negative phrase: not all is a negative DP, not
often a negative AdvP. The not of course has scope over the quantifier. Such phrases can
occur within the subject, as in [i], or before the subject, as in [ii]. The not here marks
clausal negation, so not often in [ii] triggers subject–auxiliary inversion. These phrases
are excluded from post-verbal position ([iii]) or central position ([iv]). Instead of [iii–iv]
we need verbal negation: I don’t agree with all your arguments and He doesn’t often visit
his parents.
9
A further case is in predicative AdjPs modified by so, as in %It was so not funny. This is a relatively new
construction, characteristic of the informal speech of younger speakers.
There are severe restrictions on what quantifiers take not as modifier in this construc-
tion (cf. Ch. 5, §11). Compare:
[4] i not all not every not many not much not often
ii ∗not both ∗
not each ∗
not most ∗
not some ?
not any
In clauses with verbal negation each, most, and some do not readily occur within the
scope of negation. The salient interpretation of I hadn’t read most of it, for example, is
“Most of it I hadn’t read”. There is therefore little need for a phrase that explicitly brings
them within the scope of a negative.
Both can occur readily enough inside the scope of verbal negation: I couldn’t afford
both of them. But there is still little need for a phrase combining not with both. Suppose
two swimmers have attempted to swim the Bering Straits but have not both succeeded.
If we want to express the quantification in the subject we would very likely be in a
position to say Neither of them succeeded or Only one of them succeeded, which are
more informative, and hence generally preferable to the inadmissible ∗Not both of them
succeeded.
Not any is of doubtful acceptability. Normally one would instead use no or none, as in
None of her friends had supported her, but not any is marginally acceptable as an emphatic
alternant: ?Not ANY of her friends had supported her.
(b) Not one

[5] i Not one person supported the proposal.
ii They had found not one mistake.
Not one has a somewhat wider distribution than the items in [4i], in that it can occur
post-verbally, as in [5ii]. We can also have not in combination with a or a single: They
had found not a single mistake.
(c) Not two, etc.: “less than’’

[6] i Not two years ago this company was ranked in the top ten.
ii He was here not ten minutes ago.
Not behaves differently in combination with numerals denoting numbers higher than
one than in (b) above. In the first place, not two, etc., are largely confined to measure
phrases, of time, distance, and the like. They could not replace one in such examples as
[5ii]: ∗They found not two mistakes. Secondly, not in [6] marks subclausal negation: the
clauses are positive, as evident from the lack of inversion in [i], and the reversed polarity
tags, which would be wasn’t it? and wasn’t he? The clauses in [5], by contrast are negative
(though not one or not a single . . . can also occur in measure phrases with subclausal
negation). The interpretation of not in [6] is “less than”, and this sense is found also in
such expressions as not an hour ago, not long before his death, not far from the post office,
etc., which likewise do not mark clausal negation.
(d) Not a little, not a few

[7] His speech had caused not a little confusion.
Not can also combine with the determinatives a little and a few. These have paucal
meaning, and the effect of the not is to negate the low upper bound on the quantification,
giving “a fairly large amount/number” (compare the use of no in no small achievement in

[29]). The negative thus has very narrow scope and the clause itself is positive – witness
the tag hadn’t it ?
(e) Not even, not only

[8] i a. Not even Ed approved of the plan. b. Not only Ed approved of the plan.
ii a. Not even then did he lose patience. b. ∗Not only then did he lose patience.
Not commonly combines with the focusing adverbs even and only. Not even generally
marks clausal negation: the tag for [ia] would be did he? Clause [ib], by contrast, is
positive. A tag would be rather unlikely here, but the construction readily allows a
continuation with a positive connective adjunct: Max approved of it too. In [ii] we have
inversion in [a], whereas the inversion in [iib] is ungrammatical.10 The difference between
not even and not only reflects the semantic difference: it follows from [ia] that Ed did not
approve of the plan but from [ib] that he did approve of it. (In fact [ib] presupposes, takes
for granted, that he approved of it: that is why it resists the addition of a confirmatory
tag.)
Not only marks clause negation when it functions by itself as a clause adjunct:
[Not only was the acting appalling,] the movie was far too long.
(f) Not very, not quite, etc.

[9] i We had a[not very amicable] discussion.
ii It somehow sounded [not quite right].
iii I found his story [not wholly convincing].
iv He spoke [not very confidently].
v [Not very many of them] had been damaged.
Not combines with various degree expressions that can modify adjectives, adverbs, or
certain determinatives. In [i] it combines with an attributive AdjP, in [ii–iii] predicative
AdjPs, in [iv] an AdvP, in [v] a DP. While [i–iv] illustrate subclausal negation, [v] is
different: not very many behaves like not many in marking clausal negation and being
limited to pre-verbal position.
(g) Not unattractive: not with affixally negated adjectives

[10] i Morton was in his early fifties and not unattractive to women.
ii It was a not undistinguished private university with a large endowment.
iii They had fixed the walls, and purchased some not inelegant furniture.
In general, attributive adjectives cannot be negated directly by not: cf. ∗a not large house
or ∗It looked not large. The not + adjective construction illustrated in [10] is permitted
only when the adjective consists of a base preceded by a productive and transparently
negative prefix. Note then that we cannot have ∗a not anarchic society, ∗several not intrepid
explorers: anarchic and intrepid are etymologically divisible into negative prefix + base,
10
Not only then would not be fronted even without inversion: to give it prominence we would use the cleft
construction with verbal negation: It wasn’t only then that he lost patience. Not even can occur with subclausal
negation, and thus without inversion, when the focus is a measure phrase: Not even two years ago this company
was ranked in the top ten, wasn’t it?
but this analysis is not synchronically transparent.11 A further condition is that the
adjective must be gradable. This excludes examples like ∗a not immoral purpose, ∗this not
uncrystalline substance, or ∗a not illegal act. In these uses the adjectives are classificatory
rather than gradable: purposes are either moral or not, substances either form crystals
or they do not, acts are either legal or not.
There has been occasional prescriptive condemnation of the ‘not un·’ construction,
though most manuals are perfectly clear in their view that it is fully acceptable. It would
certainly be mistaken to imagine that attractive should or could be substituted for not
unattractive to express the same meaning. The two have different meanings. The adjective
attractive denotes an appearance that ranks towards the positive end of a scale that has
ugliness at the negative end, beauty at the positive end, and a range of indeterminate
looks in the middle. The un· reverses the orientation of the scale to give an adjective
denoting an appearance ranked towards the lower end of the scale. The not yields a
negation of being towards the lower end of the scale, suggesting an appearance ranked
towards the positive end, but only guardedly so, since the middle of the range is not
excluded, and the phrase is too cautious to suggest that the user intended to indicate a
high degree of beauty:
[11] Impressionistic graph of the meanings of attractive and not unattractive
attractive
not unattractive
ugliness intermediate looks beauty
(h) Not unnaturally: negation of adverbs

[12] i Not unexpectedly, Charles was late for the meeting.
ii Not unreasonably, he asked for payment in advance.
iii Not surprisingly, they didn’t want any part of it.
Again, adverbs cannot in general be negated directly by not: cf. ∗Not stupidly, he asked for
payment in advance. The exceptions illustrated in [12] are similar to those with adjectives
in [10] – except that surprisingly (which is roughly synonymous with unexpectedly)
accepts not even though it has no negative prefix. The not in [12] has scope over just the
adverb: the clauses in [i–ii] are positive, while that in [iii] is negative by virtue of the
verbal negation.
(i) Not with PPs

[13] i Not at any stage of the proceedings did she contemplate giving up.
ii Not for the first time, she felt utterly betrayed.
Not can modify a limited range of PPs, resulting either in clausal negation as in [i],
or subclausal negation, as in [ii]. The difference here is again shown by the pres-
ence or absence of inversion. It correlates also with a difference in meaning: [i] en-
tails that she did not contemplate giving up, while [ii] entails that she did feel utterly
betrayed.
11
A subtle point here derives from the fact there are adjectives treated
⎜
as affixally negated by some speakers
and not by others. For example, impious is pronounced /im paiəs/⎜ by some, making it clear that it is
analysable into negative prefix + pious; but others pronounce it as / impiəs/, not related in pronunciation
to pious. In general, %a not impious man is acceptable for the first group of speakers, but not for the
second.
(j) Not in verbless clauses

[14] i Not an accomplished dancer, he moved rather clumsily.
ii Not under any illusions about the matter, he continued to be cautious.
iii We need someone not afraid of taking risks.
These are equivalent to the verbal constructions not being an accomplished dancer, not
being under any illusions about the matter, who is not afraid of taking risks. As in these
latter examples, the negative does not have scope beyond the subordinate clause: the
matrix clauses in [14] are all positive.
(k) Not + that clause

[15] i The film never quite generates his trademark level of icy paranoia. Not that it doesn’t
try.
ii I don’t think they should be allowed to use our public health services – not that I
have anything against immigrants, of course.
iii There are spare blankets in here, not that you’ll have any need of them.
This construction may be glossed as “This is not, however, to say/suggest that . . .”. In
each case, the not calls up a proposition that might be naturally assumed or expected in
the context, and denies that it is in fact true. In [i] what is denied is that the film doesn’t
try to generate the icy paranoia usually found in the director’s works; in [ii] that the
speaker has xenophobic views; and in [iii] that there is reason to think that it will be
useful to know where the spare blankets are kept because you might be cold enough to
need them. None of these are linguistically explicit.
The syntactic analysis is somewhat problematic. In terms of function the construc-
tion occupies a non-embedded position, like a main clause. In terms of its structure, we
might take not as modifying the content clause (as in not all it modifies all, and so on).
If so, the whole construction will have the form of a subordinate clause even though it
is not functionally subordinate; as with other cases of this kind (such as That it should
have come to this!) there is implicit rather than explicit functional subordination.
(l) Not in coordination

[16] i They are now leaving [not on Friday but on Saturday].
ii They are now leaving [on Saturday, not on Friday].
iii They’ve invited [you and your brother, but not me].
Not is found in a variety of coordinative constructions (discussed further in Ch. 15, §2.6).
In [i] not appears in the first coordinate, and there is an equivalent construction with
verbal negation: They aren’t now leaving on Friday but on Saturday. (The non-verbal
version is admissible only by virtue of the coordination: compare ∗They are now leaving
not on Friday.) In [ii–iii] not belongs in the second coordinate – without a coordinator in
[ii] (where the coordinates are conceived of as mutually exclusive), following coordinator
but in [iii]. In all of these constructions the negation is subclausal, with not having scope
over only one of the coordinates. Hence the tags or continuations seen in They are now
leaving not on Friday but on Saturday, aren’t they? and They are now leaving on Saturday,
not on Friday, and so are the Smiths.12
12
Where the not belongs in the second element, the negated phrase may be detached as a supplement, so that
what precedes is a whole clause, which may take its own tag: They are now leaving on Saturday, aren’t they, not
on Friday?
The coordinates in construction [16ii] may be verbs, but the negation is still sub-
clausal:
[17] The night turned viciously cold under a sky crowded with stars that [shone, not
twinkled,] in the diamond-clear air.
The scope of the not is just the second coordinate, so this is not a case of verbal negation
of the clause – which is why we do not have supportive do (cf. ∗shone, didn’t twinkle).
Unintegrated final not

[18] i I’m so glad those old people came to the party . . . not!
%
ii Obviously the government is going to tell us the whole truth . . . not!

%
This construction is found mainly in younger-generation speech (popularised and perhaps

originated by characters in an American television comedy sketch) but is occasionally echoed
in recent journalistic writing. As a humorous way to signal irony or insincerity, a final emphatic
not is added following a clause, retracting the assertion made. A comparable effect can be
achieved by attaching I DON ’T think instead of not.
3.2 The synthetic absolute negators

The absolute negators are:
[19] i no, none, nobody, no one, nothing, nowhere, no place (informal AmE)
ii neither, nor, never
No and none are dependent and independent forms of the determinative no, while the
other items in [i] are compounds containing no: in spite of the orthography we take no
one (“nobody”) and no place (“nowhere”) to be grammatically single words. The three
words in [ii] are transparently related to either, or, and ever.13
3.2.1 Clausal negation

In the default case the negators in [19] mark clausal negation: we look at this construction
first and then turn in §3.2.2 to the subclausal case. The examples in [20] illustrate the
clausal negation behaviour of clauses containing these items:
[20] i [Kim had done nothing about it,] and neither had Pat. [connective neither]
ii [They never replied to your letter,] did they? [positive tag]
iii In no city has she been entirely comfortable. [subject–auxiliary inversion]
Note that in [iii] we have subject–auxiliary inversion triggered by the preposing of the
PP in no city. The negative property of no percolates upwards in a way that is similar to
that in which the interrogative property percolates upwards in In which city has she been
entirely comfortable? The negative or interrogative property percolates upwards from the
determiner to the NP, and thence to the PP (see Ch. 10, §7.9).
13
There is also an archaic alternant of nothing, namely naught. But as such it is now restricted to a handful of
collocations, as in It availed him naught, It will come to naught. In AmE naught can also be a variant spelling
of nought (“zero”), which is not syntactically a negation marker; it is discussed in Ch. 5, §7.6.
§ 3.2.1 Clausal negation 813
Alternation with verbal negation constructions

Clausal negation marked by the absolute negators is generally in alternation with verbal
negation in which no is replaced by any in the case of [19i] and by either, or, and ever in
the case of [19ii]. Compare:
[21] non-verbal negation verbal negation
ia. They showed no remorse. b. They didn’t show any remorse.
iia. We liked none of them. b. We didn’t like any of them.
iiia. You did nothing about it. b. You didn’t do anything about it.
iv a. I knew neither of them. b. I didn’t know either of them.
v a. He neither knew nor cared b. He didn’t either know or care
where his children were. where his children were.
vi a. She had never felt more alone. b. She hadn’t ever felt more alone.
There are three restrictions that apply to the alternation between the non-verbal and
verbal constructions.
(a) No verbal negation counterpart with negator in clause-initial constituent
Where the negator falls within the subject or an element preceding the subject, there is
no direct verbal negation counterpart:
[22] i a. Nobody knew where Kim was. b. ∗Anybody didn’t know where Kim was.
ii a. At no stage did she complain. b. ∗At any stage she didn’t complain.
iii a. I didn’t go and neither did he. b. ∗I didn’t go and either didn’t he.
In the case of [iib/iiib], the ungrammaticality of the verbal negation construction can
be corrected by placing the element in post-verbal position: She didn’t complain at any
stage and I didn’t go and he didn’t go either.
(b) Verbal negation counterparts with no in predicative complements
Where no determines an NP in predicative complement function there may or may not
be a verbal equivalent, depending on the interpretation of the negative NP. Compare:
[23] i a. This is no place for a child. b. This isn’t any place for a child.
ii a. This is no time to give up. b. This isn’t any time to give up.
iii a. That is no way to behave. b. That isn’t any way to behave.
iv a. I’m no angel. b. ?I’m not any angel.
v a. You are no electrician. b. ?You’re not any electrician.
vi a. He’s no friend, is he? b. ?He isn’t any friend, is he?
The examples in [i–iii] show the regular relationship between the two constructions,
but in [iv–vi] the [b] versions are of questionable acceptability and in any case do not
have the same meaning as the [a] versions. It is [iv–vi] that illustrate the usual pattern
for predicative NPs with no: pairs like those in [i–iii] are largely limited to NPs with
simple and basic head nouns like place, time, and way, and subjects like this or that.
As discussed in Ch. 5, §7.8, the interpretations of the [a] examples in [iv–vi] involve
the stereotypical qualities associated with the set that the head noun denotes: angels,
electricians, and friends in this case. Thus to make [iva] true it is neither necessary nor
sufficient that I should not be a member of the set of angels; the sentence claims that I
do not have the stereotypical qualities of angels (perfect goodness, kindness, patience,
or whatever). Analogously for [va/via]. For this sort of interpretation, the [b] versions
are not appropriate paraphrases; they convey the literal meanings that the [a] meanings
lack; for example, [ivb] means “I am not any specific member of the set of angels”.
(c) Limitation on distance between verb and negator
In principle, non-verbal negators marking clausal negation can appear in any position in
the clause. However, as the position 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 late stage in the processing of the sentence:
[24] i 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.
ii 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.
When the negator is in a subordinate clause, particularly a finite one, and is signifi-
cantly far away from the matrix clause verb, it will typically not be interpreted as negating
the matrix clause, but rather will be heard as negating the subordinate clause in which
it is located. Compare, for example:
[25] i I was not trying to imply that Bob had offered bribes to any official.
ii I was trying to imply that Bob had offered bribes to no official. [= [i]]
iii I was trying to imply that Bob had not offered bribes to any official. [= [ii]]
Example [ii] is not synonymous with [i]: it is, rather, an alternant of [iii]. Compare
also:
[26] i I cannot recall actually seeing a magpie attempting to steal anything.
ii #I can recall actually seeing a magpie attempting to steal nothing. [= [i]]
In [26i] there is an occurrence of anything in a subordinate clause with primary verb
negation in the matrix clause. If the primary verb negation is removed and the anything
in the subordinate clause is replaced by nothing, we get [ii], but this is not equivalent
to [i]. Example [ii] is understood with subordinate clause negation, either of the steal
clause or the attempt clause; e.g., picking the attempt clause as the scope, the meaning is
“I can recall actually seeing a magpie not attempting to steal anything”. But seeing a bird
not attempting to steal is hardly worth mentioning, so [ii] sounds bizarre.
It is not entirely impossible for a negator in a subordinate clause to negate the matrix
one, merely unusual. Where it happens, the subordinate clause will virtually always be
non-finite, usually infinitival:
[27] i I don’t know why they say they were forced to take their shoes off; [WE certainly
forced them to do nothing of the kind].
ii [We are requiring people to pay nothing for the concert,] but nonetheless we are
hoping for at least some donations at the door.
Here the negator is located in the underlined infinitival clause but marks negation of the
bracketed finite clause, as shown by the equivalence with the verbal negation versions
W E certainly didn’t force them to do anything of the kind and We are not requiring people
to pay anything for the concert. In cases like this, then, the negative property percolates
upwards from one clause into a higher one.
§ 3.2.2 Subclausal negation 815
3.2.2 Subclausal negation

We review here three situations in which absolute negators mark subclausal negation.
(a) Negative NP as complement of preposition

[28] i I could do a lot for this place with no money at all.
ii Kim regretted having married someone with no ambition.
iii It was a matter of no consequence.
iv They were arguing about nothing.
v She finished it in no time.
The PPs in [i–iii] are semantically equivalent to clauses or PPs with clausal complements:
compare even if I had no money at all, who had no ambition, which had no importance. And
just as the negative in these latter constructions would not percolate upwards beyond the
subordinate clause, so the scope of the negative in the PP examples does not percolate
upwards into the containing clause. Example [iv] is, strictly speaking, ambiguous. It
has an interpretation as clausal negation, equivalent to the verbal negative They weren’t
arguing about anything; much more likely, however, is the meaning “They were arguing
about something completely trivial”, and here the clause is positive (with weren’t they?
as tag). In no time in [v] is an idiom meaning “extremely quickly”; it is an instance of
hyperbole: a desirably small amount of time implying a very high speed is overstated
as being zero and thus implying infinite speed. Also idiomatic is the PP in They were
obviously up to no good.14
(b) No + mean, small, etc.

[29] Getting that degree was no mean achievement, wasn’t it?
Here, in contrast to [28v], we have understatement: we understand “quite an achieve-
ment”, a positive meaning. Similarly: His resignation was in no small measure involuntary,
wasn’t it?
(c) Semantically clausal NP

[30] i They predicted no rain.
[ambiguously clausal or subclausal]
ii They promised no increase in income tax.
Example [i] has an ordinary clausal negation interpretation, equivalent to They didn’t
predict any rain. But it can also be interpreted as subclausal negation, with the NP serving
as a compressed expression of a clause, “They predicted that there would be no rain”.
Similarly for [ii].15
3.3 The approximate negators

The class of approximate negators comprises the following seven words:
[31] i determinatives: few, little
ii adverbs: rarely, seldom; barely, hardly, scarcely
14
A non-prepositional idiom is no end, meaning “very much”, as in We enjoyed it no end and so did the others,
or She had no end of a good time, didn’t she?
15
This type is not mutually exclusive with (a) above, for such NPs can occur as complement to a preposition:
The weak US dollar is expected to weigh on equity and bond markets, despite no signs of inflation in Australia
(“despite there being no signs of inflation”).
Few and little are determinatives. They function in NP structure as determiner (Few
people liked it) or fused determiner-head (Few of them liked it). Few selects count plural
heads, little non-count singulars. Little also functions as degree adjunct modifying verbs
(He little understood the implications of what he had done) or comparatives (He felt
little better). Few and little are the plain forms of the lexemes few and little, but the
comparative and superlative forms ( fewer, fewest, less, least) do not behave syntactically
as negators.
The words in [31ii] are adverbs. Rarely and seldom are adverbs of frequency, while the
other three are adverbs of degree, characteristically modifying verbs (She hardly moved),
adjectives (He was barely intelligible), and a restricted range of determinatives (especially
any: There was scarcely any food left).
We refer to these items as approximate negators on the basis of such contrasts as the
following with absolute negators or verbal negation:
[32] i a. Few of them will survive. b. None of them will survive.
ii a. Ed rarely leaves the house. b. Ed never leaves the house.
iii a. She had hardly moved. b. She hadn’t moved.
While the [b] examples indicate absolute zero, those in [a] express an imprecise quan-
tification which is close to or approximates zero. However, the fact that the approximate
negators do not indicate absolute zero gives them a somewhat equivocal status with
respect to the positive vs negative contrast. Take [iia], for example. This entails that Ed
doesn’t often leave the house, that he leaves the house no more often than occasionally,
and in this respect has a negative meaning. On the other hand, it implicates that he
sometimes does leave the house: in this respect it differs from [iib] and has to some ex-
tent a positive character. We will attempt to shed some light on the status of these forms
by considering three issues: the nature of the “not zero” implicature, their likeness to
prototypical negators with respect to what we will call the ‘direction of entailment’, and
their behaviour in the constructions we have used to distinguish between negative and
positive clauses. Few and little contrast with a few and a little, which are unequivocally
positive, and it will help clarify the status of few and little to show how they differ from
a few and a little in these three areas.
(a) Nature of the “not zero’’ implicature

As imprecise quantifiers, the approximate negators cover a range of the relevant scale of
quantification. In general, they entail that the upper bound of that range is not high but
only implicate that the lower bound is not zero. We can show the relations between a
few and few in relation to multal many and the absolute negator no in a diagram. The
arrows stand for two different relations according to the labels: ‘E∼’ means ‘entails the
falsity of ’ and ‘I∼’ means ‘implicates the falsity of ’.
[33] A few have resigned.
E˜
I˜
Many have resigned. None have resigned.
E˜
I˜
Few have resigned.
§ 3.3 The approximate negators 817
A few have resigned implicates that Many have resigned is false, and entails that None have
resigned is false. The meaning of a few sets a lower bound of three, so if no one has resigned
then A few have resigned is quite clearly false. If I know that many have resigned, then
I would not normally use the sentence with a few, since to use many would be more
informative. But the statement with a few would not be false in this context: if many
have resigned, then it is necessarily true that a few have resigned. And it could be that
I didn’t yet know whether many had: I could still make the claim with a few without
excluding the possibility that the stronger claim with many would also turn out to be
true.
Now with few we have precisely the reverse situation: Few have resigned entails that
Many have resigned is false, and implicates that None have resigned is false. The first of
these is obvious, but it is not immediately obvious that the second is just an implicature,
because it is a very strong implicature: I would be very unlikely to say Few have resigned
unless I knew that at least some had resigned. But in other cases it can be less strong.
Consider such an example as Few of you will have experienced the kind of intimidation
which our colleague Kim Jones has had to endure over the last several months. Here it could
well be that none of you have in fact experienced it: in this case I say few rather than none
not because the latter would be false but because I do not have the knowledge to justify
the stronger claim that it makes.
The difference between a few and few is reflected in constructions where the implica-
tures are cancelled (cf. Ch. 5, §5.2):
[34] i A few of them, indeed quite a lot, had found the proposal offensive.
ii Few of them, if any, will find the proposal offensive.
In [i] indeed quite a lot cancels the “not multal” implicature of a few: we could not
substitute few for a few because entailments cannot be cancelled in this way. In [ii] if
any cancels the “not zero” implicature of few, and again we could not substitute a few,
which has “not zero” as an entailment.
There may of course be other features of the clause that strengthen such “not zero”
implicatures into entailments. One such case is the common construction where barely,
hardly, or scarcely is followed by when:
[35] I had hardly arrived at the office when I was summoned to see the boss.
The meaning is that I was summoned to see the boss virtually immediately after reaching
the office. Here it is entailed, not merely implicated, that I did get to the office.
(b) Direction of entailment

The concept direction of entailment may be illustrated by comparing the entailments
in two sets of sentences, the first unproblematically negative, the second equally clearly
positive. The negative set is given in [36]:
[36] i No one would deliberately injure an animal.
ii No one would deliberately injure a mammal.
direction of entailment
iii No one would deliberately injure a horse.
iv No one would deliberately injure a racehorse. ↓
The VPs have meanings expressing successively harder conditions to satisfy: only certain
animals are mammals; only some of those are horses; and only some horses are racehorses.
Yet if [i] makes a true statement then all the sentences below it do. Assuming an ordering
where the most general is at the top and the most specific at the bottom, we can say
that the quantified NP no one is downward entailing: each example entails any that are
lower on the scale but does not entail any that are higher. If no one would deliberately
injure a mammal then no one would deliberately injure a horse, but it does not follow
from there being no one who would deliberately injure a horse that there is no one who
would deliberately injure some other kind of mammal, such as a pig.
Consider now the positive set given in:
[37] i Most people can afford to keep an animal. ↑
ii Most people can afford to keep a mammal.
direction of entailment
iii Most people can afford to keep a horse.
iv Most people can afford to keep a racehorse.
Here the entailments work in the opposite direction. If most people can afford to keep
a horse, then most people can afford to keep a mammal, but it doesn’t follow from the
fact that most people can afford to keep a mammal that they can afford to keep a horse:
there could be lots of people who can afford to keep only a small mammal such as a cat.
Here then we will say that the quantified NP is upward entailing.
Not all quantifying expressions induce one or other type of entailment. For example,
Exactly ten members of the class own a horse neither entails nor is entailed by Exactly ten
members of the class own a mammal. Where a quantifier does yield relevant entailments,
however, we can ask whether they work in a downward direction, as with negative no,
or in an upward direction, as with positive most. This test then shows few to pattern like
no, while a few behaves like most. Compare, for example:
[38] downward entailing upward entailing
i a. Few good drivers ignore signs. b. A few good drivers ignore signs.
ii a. Few good drivers ignore big signs. b. A few good drivers ignore big signs.
The predicate ignore big signs defines a more restrictive condition than ignore signs, so
from the fact that [ia] entails [iia] we know that few good drivers is a downward entailing
quantified NP. But [ib] does not entail [iib]: there could be a few good drivers who ignore
signs, though always small ones, but none who ignore big ones. It is [iib] that entails
[ib]: if there are a few who ignore big signs there must be a few (at least these same ones)
who ignore signs. So a few good drivers is an upward entailing quantified NP.
These results tie in with those obtained in considering the entailments and implica-
tures in (a) above. If few entailed “at least some” then we couldn’t have the entailment
that does in fact hold between [38ia] and [iia]: it certainly doesn’t follow from the fact
that at least some good drivers ignore signs that at least some ignore big signs. And
similarly for the [b] examples. If a few entailed “not many”, [iib] could not entail [ib],
as in fact it does. It does not follow from the fact that not many drivers ignore big signs
that the number who ignore signs of any size is also small.16
16
Downward entailment does not provide a foolproof indication that the construction is negative. This is shown
by the expression at most. At most 25% of Australians own an animal entails At most 25% of Australians own a
horse, and so on, yet these clauses (unlike the equivalent ones with no more than) are positive – as is evident
from the connective in such a continuation as and at most 25% of New Zealanders do so too/∗either, or the
tag in At most 25% of Australians own an animal, don’t they?
§ 3.3 The approximate negators 819
The direction of entailment test gives the same results for the other approximate
negators. It is illustrated for seldom in:
[39] i I had seldom seen such birds.
ii I had seldom clearly seen such birds.
iii I had seldom clearly seen such birds through binoculars.
iv I had seldom clearly seen such birds through really powerful binoculars.
The conditions imposed by the VPs of successive examples are, as before, increasingly
hard to meet (more people will have seen such birds than will have clearly seen them;
more will have clearly seen them than will have clearly seen them through binoculars;
and so on). Again, then, we have downward entailment: if [i] is true then [ii] is true; if
[i] and [ii] are true then [iii] is true; and so on downward into more and more specific
claims. This downward entailment property is what the adverb seldom contributes, for
notice that if seldom is removed the property disappears: it is not necessarily the case
that if I had seen such birds is true then I had clearly seen such birds is true and that I had
clearly seen such birds through binoculars is true, and so on.
Care should be taken when determining whether adverbs of this sort are downward entailing
to make sure that the implicit context of comparison is not changed. For example, take the
sentences in [40]:
[40] i David rarely watches films.
ii David rarely watches violent films.
It is natural to interpret [i] against a background consisting of all the days of David’s life, so
that if he watches only one film a year, [i] is true. And it is natural to interpret [ii] against a
background of all of David’s visits to the cinema, so that if 90% of the films he watches are
violent ones, [ii] is false. Now, it is perfectly possible to envisage David watching just ten films
in ten years, but with nine of the ten being violent films. In that case a natural interpretation
of [i] is true (he watches very few films per year) and a natural interpretation of [ii] is false
(in fact most of the films he sees are violent). But that is not grounds for doubting that rarely
is downward entailing.
Selection of the reference class for assessing rarity is crucial to the interpretation of the
word rare. The proper comparison is between the truth conditions of the two sentences given
a constant choice of reference class. For example, if we fix the reference class as the set of days
in David’s life, then watching films on average ten times in ten years certainly counts as rarely
watching films, so [i] is true; but by the same standard, watching violent films on average
only nine times in ten years certainly counts as rarely watching violent films, so [ii] is true
(as it would be even if all David’s film choices were violent).
This is not to say that it is illegitimate to select the set of David’s film viewing experiences
as the reference class for evaluating the truth of [40ii]; but if we are concerned with whether
[i] entails [ii], we should not shift reference class between sentences so that [i] is evaluated by
reference to one class and [ii] is evaluated by reference to another. The claim of downward
entailing is that if we select a reasonable reference class that makes [i] true, when we fix that
class and use it to evaluate [ii], [ii] will also be true.
(c) The syntactic tests for negative polarity

Examples such as the following show that the approximate negators can mark clause
negation:

[41] i a. Little of the liquid spilled, not even when the flask fell over.
[not even]
b. One seldom sees such birds, not even in Australia.

ii a. She hardly goes out these days, and neither does her son.
[connective]
b. Little of the gas spilled, and little of the gas escaped, either.

iii a. Few good drivers ignore signs, do they?
[tag]
b. Hardly any of them complained, did they?
iv Rarely does the possum emerge before dusk. [inversion]
Matters here, however, are somewhat more complex than with the absolute negators:
examples like those in [41] are not fully representative of the patterns found. There are
four points that need to be noted.
Approximate negators mark clausal negation more readily when positioned early
The negators in [41] all precede the verb. In cases where they occur late in the clause the
polarity tests often give much less clear-cut results. Compare, for example:
[42] i Few of the boys had shown any interest in the proposal.
ii He had so far shown the visitors few of the sights of London.
While ‘had they? ’ is perfectly acceptable as the tag for [i], ‘had he? ’ for [ii] is for many speakers
at best marginal. And [i] allows the continuation and nor indeed had many of the girls, while
and nor indeed had his colleagues is questionable as a continuation of [ii].
The strength of the “not zero” implicature may affect the syntactic polarity
[43] i He’s probably lying. It’s barely conceivable that he could have done it himself.
ii She’s barely alive.
A positive tag is more acceptable for [i] than for [ii], and this would seem to correlate with the
fact that [i] has a somewhat stronger negative flavour than [ii]: [i] suggests that I’m inclined
to believe he couldn’t have done it himself, while [ii] seems to be saying that she is alive,
though only just.
The class of approximate negators is not entirely homogeneous

Rarely and seldom are somewhat weaker markers of clausal negation than the others. These
two can be the focus for only, while the others cannot:
[44] i She visits her parents only rarely.
ii ∗She had read only few of the letters.
iii ∗She had done only hardly anything about it.
Note that [ii] can be corrected by replacing few by a few. The only strongly favours an
interpretation where there are at least some occasions on which she visits her parents, and
a does she ? tag for [i] or a continuation like and neither does her brother would be quite
unacceptable. It is also possible, for many speakers at least, for rarely to occur in front position
without triggering subject–auxiliary inversion: Rarely, the possum emerges before dusk. Here
we understand “very occasionally, on rare – but at least some – occasions”, and the clause is
syntactically positive.
§ 3.4 Affixal negation in relation to verbal negation 821
The approximate negators can occur in subclausal negation like the absolute ones
Some of the constructions where the absolute negators mark subclausal negation, illustrated
in §3.2.2, also permit approximate negators:
[45] i I could do a lot for this place with barely any money at all. (cf. [28i])
ii Kim regretted having married someone with little ambition. (cf. [28ii])
iii They have predicted little rain for the next month. (cf. [30])
3.4 Affixal negation in relation to verbal negation

We have noted that affixal negation is always syntactically subclausal, but from a semantic
point of view it may or may not be equivalent to clausal negation. Compare:
[46] i a. That model is available. b. Such mistakes are common.
ii a. That model is not available. b. Such mistakes are not common.
iii a. That model is unavailable. b. Such mistakes are uncommon.
Examples [iia] and [iiia] are equivalent, and both are contradictories of the positive [ia].
It is not possible for [ia] and [iia] to be both true, and it is also not possible for them to
be both false, and the same holds for [ia] and [iiia]. In the [b] examples, however, the
clausal and affixal negatives are not equivalent: while [iib] is again a contradictory of the
positive [ib], the affixal negative [iiib] is not. It is not possible for [ib] and [iiib] to be
both true, but it is possible for them to be both false – so [iiib] is the contrary of [ib].
This is evident from the fact that it makes perfect sense to say:
[47] Such mistakes are not common, but they are not uncommon either.
The difference between set [a] and set [b] in [46] is due to the fact that common
denotes a gradable property whereas available does not (cf. Ch. 6, §2.2). Common and
uncommon can be thought of as applying to non-adjacent areas on a single scale, as
shown impressionistically in [48]:
[48] not common
uncommon common
Frequencies in the middle area of the scale don’t qualify as either common or uncommon,
so not common covers a larger area of the scale than uncommon.
In practice, Such mistakes are not common will tend to be interpreted with the fre-
quency falling towards the left part of the scale. But this is a matter of implicature, which
we will take up in §5: for present purposes it is sufficient to note that examples like [47]
demonstrate that Such mistakes are not common doesn’t have the same meaning as Such
mistakes are uncommon.
The difference is accentuated if we add the intensifier very:
[49] a. Such mistakes are not very common. b. Such mistakes are very uncommon.
Very common denotes a narrower area of the scale than common, located at the right of
diagram [48], and similarly very uncommon a narrower area than uncommon, located
at the left: the in-between area is therefore greater, and the difference in meaning more
obvious.
4 Polarity-sensitive items
A significant number of items – specific words, fixed or variable phrases, and idiomatic
expressions – are polarity-sensitive, i.e. sensitive to the polarity of the environment in
which they occur. Some items are admissible in negative environments but not normally
in positive ones, while others occur in positive environments but generally not in negative
ones:
[1] i a. She doesn’t see him any longer. b. She knows him already.
ii a. ∗She sees him any longer. b. ∗She doesn’t know him already.
The aspectual adjunct any longer, for example, is acceptable in negative [ia] but not in
positive [iia]. And conversely already is acceptable in positive [ib] but not in negative
[iib] (ignoring the special case when it is used to contradict a previous utterance like
[ib]). We need two pieces of new terminology for referring to these two kinds of item:
[2] i Items which prefer negative contexts over positive ones (such as any longer) are
negatively-oriented polarity-sensitive items, or NPIs.
ii Items which prefer positive contexts over negative ones (such as already) are
positively-oriented polarity-sensitive items, or PPIs.
Non-affirmative contexts
The reference to preferring negative contexts above is not incorrect, but it is only part of
what needs to be said about the distribution of NPIs. The clauses in [1] are declarative
main clauses. When we go beyond such data to consider interrogatives and certain types
of subordinate clause we find that NPIs are not restricted to negatives:
[3] i Do you need me any longer? [interrogative]
ii If [you play any longer], you’ll get too tired. [conditional]
In [i] any longer occurs in an interrogative clause, while in [ii] it occurs in a subordinate
clause functioning as protasis in a conditional construction. Although these clauses are
positive, they have it in common with [1ia] that they are not used to assert a positive
proposition: the proposition is questioned or merely conditionally entertained. This
larger class of contexts that admit NPIs we call non-affirmative contexts – and those
that exclude them, by contrast, are affirmative contexts.
While NPIs are restricted to non-affirmative contexts, PPIs are not restricted to affir-
mative contexts. Already, for example, is just as acceptable as any longer in interrogative
and conditional constructions:
[4] i Have they already left? [interrogative]
ii If [he has already finished his work], we can leave immediately. [conditional]
Thus NPIs and PPIs do not occur in mutually exclusive sets of environments. They are
mutually exclusive in declarative main clauses, but not always elsewhere.17
17
NPIs are often referred to as ‘negative polarity items’. We avoid that term because it is open to misinter-
pretation. NPIs are not themselves negative items: the NPI any, for example, is to be distinguished from
the negative word no. Nor, as we have just emphasised, are NPIs restricted to negative contexts. Alternative
terms found in the literature for our ‘affirmative’ and ‘non-affirmative’ are ‘assertive’ and ‘non-assertive’.
§ 4.1 Negatively-oriented polarity-sensitive items (NPIs) 823
Varying strength of the restrictions

The NPI and PPI classes are by no means completely homogeneous: we find significant
differences with respect to the distributional restrictions that apply. Much, for example,
qualifies as an NPI by virtue of the unacceptability of clauses like ∗She loved him much,
but it is not wholly excluded from affirmative contexts, as is evident from examples like
She much regretted accepting their invitation. By contrast, the NPI at all (She did not love
him at all ) is completely excluded from affirmative contexts. Similarly with PPIs. We
have noted that already is quite acceptable in interrogatives, but pretty (“fairly”) is not:
examples like ?Is she pretty happy? are of doubtful acceptability.
It should also be borne in mind that in some cases the restrictions apply to one sense
of an item but not to another. Thus any is non-affirmative in the sense it has in She didn’t
make any changes or Did she make any changes?, but it also has a ‘free choice sense’ which
is not polarity-sensitive, as in Take any card, or Any changes must be approved by the
board (cf. Ch. 5, §7.5). Where it will help exposition we will use subscripts to distinguish
between different senses, referring to the NPI any as ‘anyn ’.
We will review the NPI and PPI classes in turn, and then examine correspondences
between certain sets such as any (NPI), some (PPI), and no (negator). In the final sub-
section we take up in more detail the question of what constitutes a non-affirmative
context.
4.1 Negatively-oriented polarity-sensitive items (NPIs)

The class of NPIs is quite large: it is not possible to give an exhaustive and definitive list.
The most important ones, and a sample of the less important ones, are given in [5].
[5] i The any class of items: anyn , anybodyn , any longer, any more (AmE anymore),
anyonen , anythingn , anywheren
ii Miscellaneous grammatical items (mostly functioning as adjuncts): at all,
eithern , evern , longn , much, till/until, toon , what(so)evern , yetn
iii The modal auxiliaries dare and need
iv A few lexical verbs: bother (+ infinitival), budge, faze
v A large and probably open array of idioms, including: can abide/bear/stand,
can be bothered, could care less,18 cost a bean, do a (single) thing (about . . . ),
drink/touch a drop, eat a bite/thing, give a damn/fig, have a clue, have a penny
(to one’s name) (BrE), have a red cent (AmE), hear/say a word/sound, hold a
candle to, in ages, in donkey’s years, lift a finger (to help), mind a bit, move a
muscle, say a word, see a thing, see a (living) soul, so much as (+ verb), take a
(blind) bit of notice, would hurt a fly
We prefer ‘affirmative’ as it suggests the two most important features: positive in contrast to negative (an
affirmative answer is a positive one) and declarative in contrast to interrogative (to affirm is to state, not to
ask). ‘Non-affirmative’ can also be applied to items, providing an alternative term for NPIs, one which we
have used in other chapters: in the present context we prefer NPI because we are dealing with the contrast
between the two types of polarity-sensitive item.
18
For many American speakers the expression I couldn’t care less has lost its negation and the expression is
now I could care less, still with the idiomatic meaning “I do not care at all”. For these speakers, care less is no
longer an NPI; could care less has become an idiom with a negative meaning (approximately the opposite of
its literal meaning). This is not an uncommon development; it is seen again in the development from I don’t
know beans about it “I don’t know anything about it” to I know beans about it with the same meaning.
The following examples provide matching negative and positive examples in declar-
ative main clauses for a sample of these.
∗
[6] i a. Lee didn’t budge an inch. b. Lee budged an inch.
∗
ii a. They can’t abide aniseed. b. They can abide aniseed.
∗
iii a. You needn’t come with us. b. You need come with us.
∗
iv a. He didn’t wait so much as a week. b. He waited so much as a week.
∗
v a. She hasn’t woken up yet. b. She has woken up yet.
#
vi a. I didn’t see a living soul. b. I saw a living soul.
#
vii a. Joe hasn’t lifted a finger to help. b. Joe has lifted a finger to help.
The positive versions are either ungrammatical or have a quite different and often
bizarre literal meaning. Thus while [via] means that I didn’t see anyone, [vib] can
only mean that in some way I actually saw a soul. Similarly, [viia] means that Joe did
nothing to help, but [viib] can only mean that Joe’s help consisted literally of raising a
finger.
The subscripts attached to some of the items in [5] indicate that they have other
senses in which they are not NPIs. Thus any and either have free choice senses (cf. Ch. 5,
§§7.5, 7.7); either also has a non-NPI use as a marker of coordination (She’s arriving
either on Monday or on Tuesday); ever and yet are similarly not NPIs in It will last for ever
and He is yet to announce his decision; nor is long an NPI as an adjective (It lasted a long
time).
Toon means “very”, as opposed to the ordinary too meaning “excessively”: the differ-
ence is seen in the well-known ambiguity of the testimonial writer’s I can’t recommend
her too highly (“it’s impossible to overstate her good qualities” or, with toon , “I can give
only a lukewarm recommendation”).
What(so)ever is an NPI, with the meaning “at all”, only when functioning as an
emphatic postmodifier in NP structure following any or no: There is no justification
whatsoever for his behaviour; Have you any idea whatever of its value?
The case with till and until is somewhat different. These are NPIs only when the clause
has a punctual meaning (cf. Ch. 8, §7.3):
[7] i a. We won’t leave till six o’clock. b. ∗We will leave till six o’clock.
ii a. We won’t publish it until next year. b. ∗We will publish it until next year.
iii a. We won’t stay until the end. b. We will stay until the end.
With verbs that are non-punctual and atelic, till and until are not polarity-sensitive,
as shown in [iii]. But this is not a matter of a difference in the sense of till/until: it is
simply that the polarity-sensitivity of these prepositions is limited to their occurrence
with certain types of VP.
4.1.1 NPIs vs negative idioms

It will be noted that we have not included the word not or the inflectional suffix ·n’t
in any of the NPIs. Thus we have listed long and could care less rather than not long or
couldn’t care less. The negative element is part of the context which admits the NPI, not
part of the NPI itself. There are in fact a number of idioms which do include negators
as subparts, and they behave quite differently from NPIs. Idioms of the type in question
include the underlined phrases in [8]: not half bad “quite good”, like nothing better “be
pleased”, and stop at nothing “use any means necessary”.
§ 4.1.2 Variation in the strength of the negative orientation 825
[8] i I tasted the caviar, and it was not half bad.

ii I would like nothing better than to attend the dinner.
iii George will stop at nothing to get that job.
Notice that the positive counterparts are not available, or not available with related meanings:
half bad, like something better, and stop at something are not positive counterparts to the
expressions given in [8].
These idioms contain a negator, but do not negate the clause they belong to. NPIs, by
contrast, do not contain a negator, but require that one of a certain range of properties be
present in the context in which they appear in order that it be non-affirmative: negation is
just one of the properties of this kind. The difference between the behaviour of negative-
containing idioms and that of NPIs is seen when we contrast, say, the distribution of not for
long with that of not half bad. At first they might seem parallel:
[9] i a. They laughed, but not for long. b. It was salty, but not half bad.
ii a. It wasn’t for long. b. It wasn’t half bad.
But the parallel rapidly breaks down:
#
[10] i a. They didn’t laugh loudly or for long. b. It wasn’t too salty or half bad.
#
ii a. No one laughs for long. b. Nothing was half bad.
#
iii a. Few people laughed for long. b. Few portions were half bad.
#
iv a. I doubt that he’ll laugh for long. b. I doubt that it was half bad.
#
v a. Did they laugh for long? b. Was it half bad?
#
vi a. I’ll leave if they laugh for long. b. I’ll leave if it’s half bad.
The ‘# ’ in the [b] cases indicates that the expected idiomatic meaning is not present (though
some other meaning might be possible, e.g. with half bad meaning “50% putrescent”). As
we see, for long does not have to be preceded by not or a negative auxiliary; it merely has to
be in some sort of non-affirmative context. The situation with not half bad is quite different.
Immediately before half bad must be either not or a negative auxiliary; merely having half bad
in some non-affirmative context is not sufficient for it to count as an instance of the idiom
not half bad.
The idiom like nothing better is likewise restricted to combinations involving those three
words: [8ii] can be paraphrased by There’s nothing I’d like better than to attend the dinner,
but not by I wouldn’t like anything better than to attend the dinner. The latter has only the
literal meaning that nothing (not even winning a national lottery) tops dinner attendance
on my preference list, whereas [8ii] is just a polite form of words carrying no such literal
entailment.
Stop at nothing also exists only in negative form. ?George will stop at something would have
to have a literal meaning and is hard to find any context for; [8iii] is not paraphrased by
?
George will not stop at anything to get that job. It has a tighter structural restriction than like
nothing better in that the word sequence stop at nothing (usually will stop at nothing) must be
intact, so we do not find ?There is nothing that George will stop at to get that job with the NPI
sense.
4.1.2 Variation in the strength of the negative orientation

We noted at the beginning of the discussion of polarity-sensitive items that there is
variability with respect to the strength of their positive or negative orientation. As far as
NPIs are concerned, there are two issues to be considered: (a) how strictly the restrictions
apply in negative clauses, and (b) the occurrence of these items in other kinds of non-
affirmative context, such as interrogatives.
The restriction to negative contexts in declarative main clauses

With some items things are fairly clear. Leaving aside certain special constructions dealt
with in §4.4 below, the NPI at all is completely excluded from positive declaratives. She
didn’t like it at all is admissible, but ∗She liked it at all is not, and so on. The same applies
to the any series of words once we set aside the semantically distinct free choice uses.
But for others we find that a word may be an NPI in some styles or registers and not in
others. The item much is a particularly problematic case of this sort, and long and many
have similar properties.
Contexts for much, long, and many
Much has a bewilderingly large set of distinct uses associated with different styles or
registers. Compare:
[11] i a. The new, more elaborated abstracts were much favoured among modernists.
b. Location theorists have given these matters much consideration.
c. This means much to the American tradition.
d. The design of an interlocking frame is much like a mechanical puzzle.
e. The president spent much of the weekend at his summer home on Cape Cod.
ii a. Thank you very much for the lovely flowers.
b. So much has happened that I’m not sure I can remember it all.
c. When I wear these I look too much like my dad.
d. This is much better than the other one.
e. I’ll tell you this much: I didn’t pay full price.
iii a. The lecture was very long but it didn’t really cover much.
b. I went sailing once but I didn’t enjoy it much.
c. He isn’t much of a dancer.
In general terms the examples in [i] can be said to be more characteristic of written
English, literary contexts, formal style, and the usage of older speakers. The uses illus-
trated in [ii–iii] are more likely to be found in spoken English, conversational contexts,
informal style, or the usage of younger speakers. But this is only a rough guide. It is
extraordinarily difficult to separate the constructions out in a rigorous way, because cer-
tain phrases from the literary language have become familiar sayings in everyday speech
almost as quotations.
We can be more specific about the examples in [11i], which will give a sense of what
we mean. In [ia], much is a premodifier of a past participle (in its passive use), and has
an intensificatory meaning like “greatly” or “to a high degree”. This is quite literary, but
even this turns up in casual conversation through a number of fixed or partially fixed
phrases that are in a sense borrowings from an earlier stage of the language; for example,
the phrase Much appreciated is often used with the sense “Thank you”. In [ib], much
is determiner in NP structure, and means “a lot of ”; again, this is literary, but while
We have much sugar is extremely unlikely to occur in casual conversation, one might
easily hear something like The living room is a scene of much confusion, I’m afraid in a
semi-jocular echo of the formal construction. Something similar is true for [ic], where
much is used as fused determiner-head. The phrase much like in [id] is distinctly literary
§ 4.1.2 Variation in the strength of the negative orientation 827
in that context, yet the phrase one is much like the next has been adopted as a common
turn of phrase in conversation. And much of the weekend in [ie] is slightly literary (a
lot of the weekend would be more conversational), but much of the time is familiar in
conversation.
There is an important difference between the cases in [11i] and those in [ii]. No cau-
tions about literariness or formality need apply to the latter: all of them are fully natural
in casual conversation in the contemporary language. It should be noted that much with
an adverbial modifier like very, so, or too as in [iia–c] has a far wider distribution than
much on its own. Much as a modifier of comparatives, as in [d], is very common (much
longer is analogous to very long); and so are the phrases this much and that much, as in
[e]. None of these illustrate NPIs.
That leaves the kind of use seen in [11iii]. These are the constructions in which much
is an NPI. This is particularly clear in contrasts like these:
[12] i a. I don’t enjoy sailing much. b. ∗I enjoy sailing much.
ii a. We don’t have much time. b. ∗We have much time.
iii a. Kim isn’t much of a dancer. b. ∗Kim is much of a dancer.
Much is thus certainly an NPI (and it is therefore on the list in [5]), despite the fact
that looking at a random selection of occurrences of much would make it easy to think
that one had found counterevidence to that statement. Many occurrences of much are,
indeed, in affirmative environments. But nonetheless, if we focus on (say) the use of
much as a post-verbal adjunct when occurring without its own premodifying adverb, it
is clearly an NPI, as the sharp contrast between [12ia] and [b] shows. Instead of [12ib]
we would have, in informal style, I enjoy sailing a lot / a great deal. The informal way to
express the positive counterpart of the claim in [12iia] would be not [12iib] but something
like We have plenty of time. And to express the positive counterpart of [12iiia] one would
use something like Kim is quite a dancer.
The durational adverb long – which, like much, expresses multal quantification –
exhibits similar behaviour. We have She hasn’t known him long but not ∗She has known
him long (cf. She has known him a long time); but long can occur pre-verbally in (somewhat
formal) positive clauses: I have long thought that this should be changed.
Multal many also shows some signs of negative orientation, though to a lesser
extent. The sentences Many were lost and We saw many flowers, with no negation, are
very slightly literary, whereas Not many were lost and We didn’t see many flowers, with
negation, are as common in casual conversation as in writing. But again, an example
like ?I’m not hungry because I’ve eaten many biscuits sounds completely unnatural,
with a lot of biscuits very strongly preferred in informal style.
Occurrence in contexts that are non-affirmative but positive

Negatives constitute the most central type of non-affirmative context, but we have noted
that NPIs can occur in certain types of positive clause, such as interrogatives – this is
precisely why we talk in terms of non-affirmative contexts rather than simply negative
ones. These positive non-affirmative contexts all have various semantic or pragmatic
links to negatives. There are, however, differences among the NPIs with respect to how
strong or close these links need to be. Such differences can be illustrated by considering
the non-affirmative context provided by closed interrogatives.
NPIs in closed interrogatives
Other things being equal, questions expressed by positive closed interrogatives are neutral
as opposed to biased (cf. Ch. 10, §4.7). That is, they do not indicate any predisposition
on the part of the questioner to think that one answer is more likely than the other. Now
the inclusion of certain NPIs in the question has no effect on this neutrality, but with
others bias towards a negative answer is introduced. Compare:
[13] i a. Did they have a dog? b. Did they have any money?
ii a. Has she been to Paris? b. Has she ever been to Paris?
iii a. Did they help him? b. Did they lift a finger to help him?
The possible answers to [ia] may be expressed as Yes they did have a dog and No they
didn’t have a dog: as we have said, there is no indication that one or other of these is
expected to be the right answer. The answers to [ib] (omitting henceforth the yes and
no) are They had some money and They didn’t have any money, and again the question
is quite neutral. The inclusion of any in the question does not indicate that the answer
containing any is expected. The absence of any indication that a positive answer is
favoured is sufficient to license the NPI any: this is the default existential quantifier for
polar questions. Similarly, both questions in [ii] are neutral as between the answers She
has been to Paris (at some time) and She hasn’t (ever) been to Paris. This is not so in [iii].
While [iiia] is of course neutral, [iiib] is not: it is oriented towards the negative answer
They didn’t lift a finger to help him. The form of the question is such that the positive
answer would be #They lifted a finger to help him, but since this (in the sense “They helped
him”) is anomalous, the question doesn’t cater directly for a positive answer and hence
indicates bias towards a negative one. The same applies to most of the other items in
[5iv–v].
Pragmatic nature of strength contrasts
It is important to emphasise that the difference between (say) any or ever on the one
hand and budge or lift a finger or in ages on the other is not a syntactic one. We are
not saying that the former can occur in positive interrogatives while the latter cannot.
The difference, rather, is pragmatic, having to do with the conditions under which they
can be used in the expression of questions and with the interpretation of the resulting
questions.
Take, for example, the phrase in ages, as in I haven’t tasted truffles in ages. This is not
easily contextualised in an interrogative: ?Have you tasted truffles in ages? would not be
a highly appropriate way of simply asking someone whether it had been a very long
time since they last tasted truffles. However, it might well be found in a context that
was strongly biased in a direction suggesting a negative answer, as in [14i]. Similarly, the
occurrence of the verb budge is almost entirely limited to negative clauses but one might
encounter it in a biased context like [14ii].
[14] i I don’t think you know anything about truffles or any other gourmet foods. Do you
eat truffles regularly? Have you even so much as tasted truffles in ages?
ii Did you budge an inch to let anyone else sit down? Not you; you just sat there hogging
the whole couch as usual.
§ 4.2 Positively-oriented polarity-sensitive items (PPIs) 829
4.2 Positively-oriented polarity-sensitive items (PPIs)

The class of PPIs is somewhat smaller than that of NPIs. The main members are listed in
[15], while [16] provides a sample of positive examples contrasted with their inadmissible
negative counterparts:
[15] i some, somebody, someone, something, somewhere, somehow, somewhat
ii the paucal determinatives a few, a little, several, various
iii the degree adverbs pretty, fairly, quitep , farp
iv aspectual already, still
v connective so, too, as well
vi the modal idioms would rather, would sooner, would as soon
vii illocutionary I guess
[16] i a. They made some mistakes b. ∗They didn’t make some mistakes.
ii a. It’s pretty big. b. ∗It isn’t pretty big.
iii a. It is still a mystery why he ran off. b. ∗It isn’t still a mystery why he ran off.
iv a. Kim saw it, and so did Pat. b. ∗Kim saw it, but not so did Pat.
v a. This one is far better. b. ∗This one isn’t far better.
vi a. I would rather die. b. ∗I wouldn’t rather die.
vii a. I guess I agree. b. ∗I don’t guess I agree.
As a PPI quite means “fairly”: with the sense “completely” quite is not polarity-sensitive
(cf. That’s quite right ∼ That’s not quite right). The far that acts as a PPI modifies
comparative expressions for the most part, and means roughly “considerably”; as a
measure of distance it is not polarity sensitive (cf. They live / don’t live far away). I guess
qualifies as a PPI where it serves to indicate the illocutionary force of the utterance: it is
not polarity-sensitive when guess has its literal meaning (as in I don’t guess what’s in my
presents, I wait to be surprised when I open them on Christmas Day).
Scope of negation
The negatives in the [b] examples in [16] have scope over the PPI. In general, PPIs are
admissible with negatives over which they themselves have scope: this does not count as
a non-affirmative context for the item concerned. Compare, for example:
[17] i I would rather not commit myself.
ii Far more of them didn’t understand it than did.
iii We already can’t afford any luxuries: how will we deal with this new expense?
iv I still don’t know how she did it.
v I didn’t understand some of the points she was trying to make.
vi I’m afraid I couldn’t stand several of her friends.
In [i] the not belongs in the subordinate clause: it modifies commit, not would rather.
Example [ii] says that the number of those who didn’t understand was far greater than
the number of those who did. Example [iii] says that the state of affairs wherein we can’t
afford luxuries already obtains. In [iv] the state of my not knowing how she did it still
obtains. And in [v] some of the points she was trying to make had the property that
I didn’t understand them. Similarly [vi] says there were several of her friends whom I
couldn’t stand, not that the number whom I could stand was less than several. In all of
these, then, the PPI itself is not affected by the negation and hence perfectly admissible.
The connective adjuncts so, too, and as well, however, cannot even take scope over
negation:
[18] i ∗Kim didn’t see it, and so didn’t Pat.
ii ∗Kim didn’t see it, and Pat didn’t see it too / as well.
This of course is why we were able to include these items in our tests for clause polarity.
There is, however, an appreciable difference in the strength of the restriction: [i] is
completely ungrammatical, whereas [ii] might be regarded as marginally acceptable by
some speakers provided the final adjunct is prosodically set off to some extent.
Metalinguistic and other overrides

A number of the PPIs can occur within the scope of negation when used in a metalin-
guistic or quasi-metalinguistic way:
[19] i A: It’s a pretty big fish. B: It isn’t pretty big, it’s absolutely gigantic.
ii A: He’s pretty stupid. B: He’s not pretty stupid, he’s actually quite bright.
In [i] we have a case of metalinguistic negation. B is not saying that the proposition “It is
pretty big” is false, but is rejecting it on the grounds that it doesn’t go far enough. In [ii],
B is rejecting A’s statement as false, so this is not metalinguistic negation; nevertheless
B is repeating A’s expression. We refer to this latter case as a denial, a contradiction of
something that has been said and explicit assertion of an alternative to it. In talking about
the unacceptability of PPIs in negative contexts we are setting aside such special uses;
and the asterisk attached to the [b] examples in [16] is to be understood as applying to
the ordinary use of the clauses in question.
PPIs may also be found in negative clauses that are embedded beneath a superordinate
negative in such a way that the positive counterpart is implicated:
[20] i You can’t tell me that it isn’t far better for some couples to divorce than to stay
together.
ii Never think that I wouldn’t rather be with you than at the office.
You can’t tell me that it isn’t in [i] strongly implicates “it is”, which sanctions the positively-
oriented far; and in [ii], never think that I wouldn’t strongly implicates I would, which
sanctions the PPI would rather.
Interrogatives
Most PPIs can occur in interrogative clauses:
[21] negative interrogative positive interrogative
i a. Wouldn’t you rather stay here? b. Would you rather stay here?
ii a. Didn’t Kim see it too / as well? b. Did Kim see it too / as well?
iii a. Isn’t he pretty stupid? b. ?Is he pretty stupid?
iv a. Aren’t they still/already in London? b. Are they still/already in London?
v a. Isn’t there something wrong with it? b. Is there something wrong with it?
Negative interrogatives are normally used to ask biased rather than neutral questions,
and when there is bias towards a positive answer this is sufficient to admit the PPIs
even though they are in the scope of the negative. Illocutionary I guess and connective
so, however, are restricted to declaratives. For the rest, the degree adverbs pretty, fairly,
quitep occur significantly less readily in positive interrogatives than the others.
§ 4.3 Correspondences between PPIs, NPIs, and negators 831
4.3 Correspondences between PPIs, NPIs, and negators

There is an important relationship between some, anyn , and no illustrated in such
sets as:
[22] i Kim made some mistakes. [positive]
ii Kim didn’t make any mistakes. [negative with verbal negation]
iii Kim made no mistakes. [negative with non-verbal negation]
In [ii–iii] the existential quantification falls within the scope of the negation, and (with
the provisos mentioned in §4.2 above) some cannot occur in this environment. Thus
[ii–iii] are negations of [i], differing in that the negation is expressed separately from the
quantification in [ii] but not in [iii].
Broadly similar relationships are found with a number of other items, so that we can
set up the following correspondences:
[23] PPIS NPIS absolute negators
i 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
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
These three displays need to be interpreted with some caution. In the first place,
the relation between paired PPIs and NPIs is not identical in all cases. For example,
the connection between somewhat and at all is considerably weaker and less systematic
than that between some and any. Secondly it should be borne in mind that a number
of the forms have a range of senses or uses, with the correspondences applying only
to some of them. Take some, for example. Of the several senses distinguished in Ch. 5,
§7.5, the connection with any and no is clearest in the basic existential quantificational
sense illustrated in [22i]. But for certain uses, most clearly that seen in That was SOME
journey (“a remarkable journey”) or SOME friend he was!, there are no close negative
counterparts.
A good number of polarity-sensitive items involve quantification. Broadly speaking,
we find that the following generalisations hold:
[24] i Universal quantifiers are not polarity-sensitive.
ii Some multal quantifiers exhibit some degree of negative orientation.
iii Paucal quantifiers tend to have positive orientation (like a few) or to be approx-
imate negators (few).
iv The most central existential quantifiers enter into the pattern of correspondences
illustrated in [23i].
Note in this connection that those uses of some that go beyond simple existential quantifi-
cation to indicate a fairly considerable quantity do not enter into such correspondences.
It took some time to rectify the problem, for example, does not have a straightforward
negative with any or no: #It didn’t take any time to rectify the problem; a pragmatically
closer negation here would be It didn’t take much time to rectify the problem.
A further general point to be made is that certain of the NPIs have free choice senses
in addition to their NPI senses. This applies to any and its compounds and to either
when it functions as determiner or fused determiner-head in NP structure: Take any
of the computers; Ask anyone; You can have either of the printers. Ever has a free choice
sense when compounded with relative and interrogative words (Take whatever you like or
He’ll grumble whatever you do). Elsewhere ever can express universal quantification: She
had been ill ever since she returned from Paris (“all the time”); Ever the optimist, she was
undeterred by these warnings. Free choice and universal quantification are often related
by implicature: Anyone can do that implicates “Everyone can do that” (see Ch. 5, §7.5).
We will now provide some further brief comments on the sets given in [23]:
(a) The some ∼ any series ([23i])

In positive interrogatives the some and any words generally contrast:
[25] i a. Did Kim make some mistakes? b. Did you tell someone? [PPI]
ii a. Did Kim make any mistakes? b. Did you tell anyone? [NPI]
The positive orientation of some is reflected in the fact that while the versions in [ii]
are quite neutral, those in [i] suggest a somewhat greater inclination on the part of the
speaker to think that the answer may well be positive.
The variants in [23id] containing place rather than where belong to AmE (informal
style).
Ever and never can function as adjuncts of frequency or temporal location. In the
frequency sense they correspond straightforwardly to sometimes: He sometimes loses
his patience ∼ He doesn’t ever lose his patience ∼ He never loses his patience.19 For the
temporal location sense with past time the closest PPI is once: She once liked them ∼
She didn’t ever like them ∼ She never liked them. For future time the PPI sometime
can be used: I hope they will change these rules sometime ∼ . . . won’t ever / will never
change these rules. AmE anytime most often has a free choice sense (Feel free to call me
anytime), but it is found as an NPI in combination with soon and suchlike expressions
as counterpart to sometime: I expect it to end sometime soon ∼ %I don’t expect it to end
anytime soon.
Correspondence between somewhat and at all is seen in pairs like I somewhat
regret agreeing to take part ∼ I don’t at all regret agreeing to take part. Somewhat is
slightly formal in style, and at all occurs in a much wider range of syntactic construc-
tions. For example, there is no positive counterpart with somewhat for He hasn’t worked
at all this week. At all commonly occurs as a reinforcing postmodifier in the structure
of NPs containing any or no, as in I hadn’t had any food at all or I’d had no food at all:
19
Never is likely to be preferred over verbal negation when ever immediately follows not or the negative verb;
a more natural use of frequency ever is seen in I don’t think he ever loses his patience. Ever can be used in
juxtaposition to never with emphatic effect; this device is associated primarily with informal conversation,
but can be found in other styles too: Any risk of a prime minister abusing this power is effectively
eliminated because he can never, ever put a political crony into the job.
§ 4.3 Correspondences between PPIs, NPIs, and negators 833
here the positive is simply I’d had some food. As a modifier of comparatives, somewhat
can also be paired with any: It was somewhat better than last time ∼ It wasn’t at all / any
better than last time.
The PPI somehow generally indicates manner (They had somehow lost their way), but
can also have a reason sense (in I somehow couldn’t understand what he was getting at
the word somehow means something like “for some unknown reason” or “in some way
that I cannot quite specify”). Anyhow exists, but not as the NPI counterpart, so this pair
is not included in [23i]. As a manner adverb, anyhow is based on free choice any but
with specialisation of meaning (They had stacked everything into the cupboard anyhow,
“without care, haphazardly”). More often it has a concessive or connective meaning, like
anyway.
(b) The aspectuals ([23ii])

The correspondence between the aspectual PPIs and NPIs is illustrated in:
[26] i a. Ed still lives with us. b. Ed doesn’t live with us any more/longer.
ii a. Jill has already finished. b. Jill hasn’t finished yet.
In the aspectual sense any more is generally written anymore in AmE, and that spelling
is spreading to BrE.20 In positive interrogatives, some speakers allow still but not any
more /longer in such examples as Does he still live with you? ∼ %Does he live with you
any more? A further difference is that any more /longer can be used for projection into
the future in a way that has no counterpart with still. I’m not working here any more
is ambiguous between a present time sense (“I no longer work here”) and a futurate
reading (“I don’t intend to continue working here”), but I’m still working here has only
the present time interpretation.
(c) The either ∼ neither series ([23iii])

One use of either and neither is as connective adjuncts in the constructions we have used
as a test of negative polarity for clauses:
[27] i a. Kim enjoyed it and so did Pat. b. Kim didn’t enjoy it and neither did Pat.
ii a. Kim enjoyed it and Pat did too. b. Kim didn’t enjoy it and Pat didn’t either.
Here neither is always in front position in a declarative clause, which excludes the possi-
bility of a direct counterpart with verbal negation + the NPI either; the matching PPI is
so. Either occurs in post-verbal position, matching too or as well. In this connective use,
nor is a variant of neither, but has a slightly wider distribution (cf. Ch. 15, §2.4).
The either and neither of [23iiic] are the ones that function as existential quantifiers
in NP structure: He hadn’t read either of them ∼ He had read neither of them. There is
here no simple PPI counterpart; we can relate these to positive He had read one of them,
but this differs from either/neither in that it doesn’t indicate selection from a set of two,
and one, moreover, is not polarity-sensitive. A closer match is with He had read one or
other of them, where the complex expression one or other is a PPI. The items in [23iiid]
are markers of coordination: this is the one place where we have absolute negators with
non-negative counterparts that are not polarity-sensitive.
20
The single word spelling is occasionally found in non-aspectual uses but this is not accepted as standard:
∗We don’t know anymore than the others. Regional AmE has a non-NPI use of anymore meaning roughly
“nowadays”: %They’re working together anymore.
4.4 Non-affirmative contexts

We review in this section the constructions or lexical items that create contexts within
which NPIs can occur. Negation is the most central of them, while the others have various
kinds of semantic or pragmatic connection with negation. In the examples that follow
we use single underlining to pick out the NPI and double underlining for the item which
sanctions it.
Negators
All negators, whether expressing clausal or subclausal negation, sanction NPIs:
[28] i a. Kim didn’t do anything wrong.
b. No one did anything wrong. [clausal negation]
c. Hardly anyone liked it at all.
ii a. He seems not very interested in any of these activities.
b. It was a matter of little consequence for any of us. [subclausal negation]
c. It is unlikely anyone has noticed it yet.
Recall, however, that the negative context begins at the point where the negator is located.
An NPI is not sanctioned by a following negator: cf. ∗Anyone did nothing wrong or ∗We
had given anyone nothing.
Interrogative clauses
We have observed that NPIs are commonly found in closed interrogatives, and for these
nothing need be added to what has been said above. Positive open interrogatives, however,
require some further discussion.
Open interrogatives
The way NPIs interact with positive open interrogatives is slightly different from the way
they work in the semantically simpler closed interrogatives. Compare the following:
[29] i a. Who helped her? b. Who did anything to help her?
ii a. Why did you help someone b. Why would you lift a finger to help
like George? someone like George?
∗
iii a. How come you like her? b. How come you like her much?
In general, questions expressed by means of positive open interrogatives have positive
presuppositions. When I ask [ia] I normally take it for granted that someone helped her
and aim to find out who it was. Similarly [iia] presupposes that you helped George and
[iiia] that you like her.
Introduction of an NPI, even one like any or ever, changes matters in various ways.
It is not presupposed by [29ib] that someone did something to help her. It may suggest
that no one did, though it doesn’t need to – cf. Who has any suggestions to make?, which
leaves it entirely open whether anyone does in fact have any.
In [29iib], with the more strongly negatively-oriented lift a finger, there is a negative
implicature – that George doesn’t deserve help, so that a proper stance would be for you
not to lift a finger to help him. In effect, the question asks for a reason for departing
from that presumption.
§ 4.4 Non-affirmative contexts 835
Example [29iiib] is ungrammatical, like the positive declarative ∗You like her much.
(Note the contrast with the fully acceptable How come you like her so much?) But again,
there is no question of much being syntactically excluded from positive open interrog-
atives: we merely need to find contexts with an appropriate negative flavour, as in Who
cares much about it, anyway?
Bare infinitival why interrogatives
[30] a. Why tell them anything about it? b. ∗Why not tell them anything about it?
It might initially seem surprising that the NPI is permitted in the positive question [a] but
excluded from the negative [b]. But again the explanation has to do with the conveyed
meaning of the construction.
Although positive, [30a] conveys the negative suggestion that there is no reason to
tell them anything about it. The negative meaning of the clause that would express this
negative implicature allows naturally for NPIs. With why not interrogatives, on the other
hand, there is a positive implicature. The conventional meaning of Why not tell them
about it? is to suggest via a rhetorical question that you should tell them about it. The
positive sense of the latter makes the NPI in [b] unacceptable.
Covertly negative lexical items with clausal or clause-like complements

Many verbs and adjectives that take clauses as complements are covertly negative in that
they trigger entailments or implicatures involving the negation of the subordinate clause,
and this is sufficient to sanction NPIs in those clauses. Certain of these items allow the
complement to take the form of an NP or PP with an interpretation like that of a clause
with NPIs still being sanctioned. In some cases the subordinate clause (or its equivalent)
may be subject: this is one case where an NPI can precede the item that sanctions it, as
in [34ii] and [36ii] below.
We group these covertly negative items into six classes: (a) failure, avoidance, and
omission; (b) prevention and prohibition; (c) denial; (d) doubt; (e) counter-expectation;
and (f) unfavourable evaluation.
(a) Expressions of failure, avoidance, and omission
NPIs are sanctioned in the complements (usually infinitival or gerund-participial com-
plements) of lexical items (mostly verbs) expressing failure to do something or similar
acts of avoidance or omission. Relevant verbs include avoid, decline, fail, forget, neglect,
refrain, and a few others, e.g. omit in BrE.
[31] i The authorities failed to do a thing to ensure the child’s safety.
ii Lee forgot to take a blind bit of notice when they were giving directions.
iii I want you to refrain from moving a muscle until you’re completely recovered.
iv We managed to avoid any further delays.
Example [i] entails that the authorities didn’t do a thing to ensure the child’s safety. In
[iv] the verb avoid has an NP as complement, understood roughly as “having any further
delays”.
(b) Expressions of prevention and prohibition
Lexical items expressing prevention, prohibition, banning, excluding, or otherwise stop-
ping actions generally take gerund-participle complements in PPs headed by from, and
NPIs are sanctioned in those complements. The verbs include ban, hinder, keep, prevent,
prohibit, stop.
[32] i We kept him from telephoning anyone before the police arrived.
ii I am prohibited from so much as naming any of the principals in this case.
Again we have negative entailments: it follows from [i], for example, that he didn’t
telephone anyone before the police arrived.
(c) Expressions of denial
The verb deny is the central item here, but a close paraphrase will work the same way
regardless of lexical content or syntactic structure.
[33] i My client denies that he ever said any such thing.
ii My client denies any involvement in the matter.
iii My client completely rejects the notion that he ever said any such thing.
Here we can paraphrase [i] as My client says that he never said any such thing. Example
[ii] illustrates the case where the complement has the form of an NP: we understand
“that he was in any way involved in the matter”.
(d) Expressions of doubt
The class of dubitative items includes verbs (doubt being the clearest case), adjectives
(doubtful, dubious, sceptical ), and nouns (doubt, scepticism):
[34] i I doubt that Lee has been to the theatre in ages.
ii That they will ever have a better opportunity is very much to be doubted.
iii I’m doubtful about the value of pursuing the matter any further.
iv She expressed scepticism about there being any point in continuing.
There is a clear relation between doubt and negation: to doubt is to entertain the possi-
bility that some proposition is false. The verb doubt with a declarative clause complement
suggests an inclination to believe that the proposition is false (see Ch. 11, §5.3.3).
(e) Expressions of counter-expectation
A statement like I’m surprised the car started asserts that I have experienced a reaction to
the discovery that the car started because that is counter to expectation: it implicates that
I had a prior expectation that could be expressed as The car won’t start. Such implicatures
are enough to sanction NPIs in complements of such verbs or verbal idioms as amaze,
astonish, astound, bowl over, flabbergast, shock, surprise, take aback, and corresponding
adjectives:
[35] i It astounds me that they took any notice of him.
ii It’s surprising he lifted a finger, considering that he’s a total stranger.
iii We were all amazed that he had been able to write anything during that time.
(f) Expressions of unfavourable evaluation
A large array of lexical items expressing unfavourable evaluations, e.g. absurd, excessive,
foolish, monstrous, ridiculous, silly, stupid, unacceptable, unwise, and many others, are
capable of providing contexts for NPIs:
[36] i It would be foolish to take any unnecessary risks.
ii Any more pudding would be quite excessive.
iii It was stupid of Basil ever to have mentioned the war.
The implicature in [i] here is that we (or whoever) should not take any more risks than
the absolute minimum, and [ii] implicates that we (or whoever) should not have any
§ 4.4 Non-affirmative contexts 837
more pudding. The situation concerned will generally be in the domain of the potential
rather than the actual: we would not say ?It was foolish to take any unnecessary risks in
speaking of a single event in the past.
Downward entailing quantified NPs

All quantified NPs that are downward entailing in the sense explained in §3.3 sanction
following NPIs. Compare:
[37] i a. Few of the bees stung anyone. b. ∗A few of the bees stung anyone.
ii a. At most ten students did any work. b. ∗At least ten students did any work.
Most downward entailing quantified NPs are negative and hence are already covered
under (a) above, but we noted in footnote 15 that at most triggers downward entail-
ment although it is not negative. As seen in [iia], it nevertheless sanctions following
NPIs. A few and at least are upward entailing, and do not create non-affirmative con-
texts.
Although [37iia] is not itself negative, it has clear links with negation. At most ten
of the thirty students worked entails that the other twenty (the majority) did not work:
compare the negative paraphrase No more than ten of the students worked, did they?
At least ten of the thirty students worked, on the other hand, has no such negative
entailment: in fact it suggests that some of the other twenty may have worked. At least
can be paraphrased by no less than, but this yields a positive clause: No less than ten of
the students worked, didn’t they?
The degree adverb too

[38] i By that time I was just too tired to budge.
ii It was too difficult for anyone else.
Too in the sense “excessively” licenses an indirect complement with the form of an
infinitival clause or a for PP. The too + positive infinitival construction can be paraphrased
by so + negative finite clause: I was so tired that I couldn’t budge. Similarly [ii] conveys
“It was so difficult that no one else could do it / that it was not appropriate for anyone
else”. Note that He wasn’t too tired to have a couple of games of tennis implicates that he
did play, and this positive implicature excludes an NPI (ignoring again the denial use):
∗
I wasn’t too tired to budge.
Prepositions (against, before, without)

Certain prepositions define contexts for NPIs, as illustrated in the [a] examples of [39]:
[39] i a. She did it without any difficulty. b. ∗She did it with any difficulty.
ii a. He left before anyone noticed it. b. ∗He left after anyone noticed it.
iii a. I argued against taking any more. b. ∗I argued in favour of taking any more.
Again the NPIs are admissible because the clauses convey negative propositions: that
she had no difficulty in doing it, that no one had noticed it, that I argued we (or
whoever) should not take any more. No such negative propositions are conveyed by the
[b] examples, which contain prepositions of opposite meaning: the NPIs are therefore
inadmissible. Without and before sanction NPIs quite generally, but against does so only
in the rather special sense it has here.
Only
[40] i Only then did she realise she had any chance of winning.
ii She remained the only one capable of making any sense of it.
Again, there is a very clear connection between only and negation. Example [i] entails
that until then she hadn’t realised she had any chance of winning, while [ii] entails that
no one else was capable of making any sense of it. For further discussion, see Ch. 6, §7.3.
Comparative and superlative constructions

NPIs are commonly found in the complement of comparative than or as, and in relative
clauses and partitive PPs in construction with superlatives. As usual, this reflects the
close association between these constructions and negatives:
[41] i She ran faster than she had ever run before.
ii The performance was as good as any you could hope to see.
iii It was the biggest fish I had ever seen.
Example [i] entails that she had never previously run as fast; [ii] entails that no perfor-
mance you could hope to see would be better than this one; and [iii] entails that I had
never before seen a bigger fish.
Overt and covert conditionals

A clause embedded as complement of if generally allows NPIs, and so do clauses in
various other constructions that have conditional interpretations. Consider:
[42] i If you want anything, just call.
ii If I’d ever seen anything like that, I’d have reported it.
iii I would read your review, if I gave a damn about your opinion.
iv Drink any more and you’ll have to get a taxi home.
Example [i] is an open conditional; it envisages that you may or may not want something,
without bias in favour of one or other of those possibilities. The mere absence of positive
affirmation is sufficient to allow an NPI such as any. The remote conditional construction
illustrated in [ii–iii] is more closely related to negation: [ii] implicates that I hadn’t ever
seen anything like that, and [iii] that I don’t give a damn about your opinion. Give a damn
is one of the more strongly negatively-oriented items, and occurs much more naturally
in a remote conditional like [iii] than an open one. In [iv] an NPI is sanctioned by the
conditional meaning even when there is no syntactically conditional construction: we
interpret the imperative as equivalent to if you drink any more (cf. Ch. 10, §9.5).
5 Increased specificity of negation (I don’t want to hear about it)
In various ways, negative clauses are often interpreted with increased specificity: they are
taken to be making a stronger claim than they actually entail. One common instance of
this phenomenon was discussed in §1.3.3, where we were examining the effect of selecting
as focus an element falling within the scope of the negation. Thus YOUR children don’t hate
school is interpreted as saying not just that the conditions for the truth of the proposition
“Your children hate school” are not all satisfied, but as indicating, more specifically, that
it is the condition associated with your that is not satisfied.
§5 Increased specificity of negation 839
Another common case of this general phenomenon is illustrated in:

[1] i a. Mary doesn’t like you. b. Mary dislikes you.
ii a. He doesn’t have many friends. b. He has few friends.
iii a. The weather wasn’t very good. b. The weather was rather poor.
In each pair, [a] implicates [b], which has a more specific meaning.
Consider [1ia]. This would presumably be true if Mary had never even heard of you,
and in this case [ib] would be false. But it would normally be quite pointless to say [ia]
if this were the case, so this scenario can be ignored in interpreting it. It would also be
true if Mary did know you but hadn’t formed a judgement about you or had no feelings
about you, positive or negative: in this case she neither likes you nor dislikes you, so [ia]
is true and [ib] is false. However, if this were so, I would normally be expected to say so
explicitly: Mary doesn’t like you, but she doesn’t dislike you either. Much the most likely
scenario for [ia] is one where Mary dislikes you, and in the absence of indications to the
contrary that is how it will generally be interpreted. The interpretation is more specific
than the actual meaning of [ia] in that it ignores the various other conditions under
which [ia] could be true.
Many in [1iia] is an imprecise multal quantifier: there’s no clear lower bound such
that twenty, say, would count as ‘many’ but nineteen would not. The scenario in which
I can be confident in saying [iia], therefore, is one in which the number of friends
falls well below what would count as many – i.e. one in which he has few friends.
So [iia] will generally be taken as conveying that he has few friends. But again that
is not what it means: it makes perfect sense to say He doesn’t have MANY friends, but
he has a reasonable number.21 The implicature [iib] is thus more specific than [iia],
excluding the middle ground on the numerical scale that lies between few and
many.
Very in [1iiia] likewise has an imprecise multal meaning, and again the middle ground
on the good–bad scale is discarded, so that [iiia] implicates [iiib]. It is easy to see that
the meanings are different, for very good denotes a relatively small part of the scale
at the top end, so that [iiia] places the weather within the large area of the scale below
that top segment, whereas [iiib] places it towards the bottom end. The implicature can
be cancelled, typically by using contrastive stress on very: It wasn’t VERY good, but it was
quite reasonable. This kind of implicature is found quite generally with multal quantifi-
cation. A third example is provided by She doesn’t often lose her temper, which implicates,
but does not entail, She rarely loses her temper.
Our major concern in this section is with a further special case of this tendency to move
to a more specific interpretation of a negative clause: the case where a clause containing
a clausal complement has negation in the matrix clause interpreted as applying to the
subordinate clause, as in the [a] examples of:
[2] matrix negation subordinate negation
i a. I don’t want to hear about it. b. I want to not hear about it.
ii a. Mary didn’t want you to tell them. b. Mary wanted you not to tell them.
21
Where not combines with many to form a DP not many, the paucal interpretation is an entailment, not an
implicature: not many means “few”. Note then that we can’t say #Not many people came to the meeting but a
reasonable number did.
In both pairs the negative marker is located in the topmost clause in [a], but in the
complement of want in [b]. And in both cases [b] is an implicature of [a] with a somewhat
more specific meaning.
This case is very similar to [1ia] above. Example [2ia] is true if either I have no feelings
about the matter one way or the other, or I want to not hear about it (to be spared from
hearing about it). In practice, the first of these possibilities is normally discounted, giving
the more specific interpretation where it is the second condition that obtains – i.e. the
interpretation expressed by [ib]. If I had no feelings about the matter, I would typically
be expected to say so: e.g. by saying I don’t mind whether I hear about it or not. But the
fact that the truth of the latter would be sufficient to make [ia] true shows that [ia] does
not have the same meaning as [ib]. This is why [ib] is an implicature of [ia], not an
entailment. And again the implicature can be cancelled: I don’t want to hear about it and
I don’t want to not hear about it – I’m completely indifferent.
Matrix verbs and adjectives allowing the subordinate negation implicature

The phenomenon illustrated in [2] is quite widespread: there are a good number of
lexemes which behave like want in constructions containing a non-finite or finite sub-
ordinate clause. In [3] we list a sample of the verbs and adjectives concerned, classified
into broad semantic groups:
[3] i wanting: choose, intend, mean, plan, want
ii advice: advisable, advise, had better, be meant, recommend, ?suggest, be
supposed
iii probability: likely, ?probable
%
iv opinion: anticipate, believe, expect, feel, %figure, guess, imagine,
reckon, suppose, think
v perception: appear, feel, look, sound, seem
The common semantic factor: ‘medium strength’
The items in [3] all have to do with various kinds of modality, and they have it in
common that their value on the dimension or scale we have called ‘strength’ is medium,
as opposed to weak or strong: see Ch. 3, §9.2.1. Consider again want, which contrasts
with weak willing and strong insist. Compare:
[4] i a. I’m not willing to be included. b. I’m willing to not be included. [weak]
ii a. I don’t want to be included. b. I want to not be included. [medium]
iii a. I don’t insist on being included. b. I insist on not being included. [strong]
It is clear that [b] is an implicature of [a] only in case [ii]. In [i], [a] actually entails
[b], but [a] is much more informative than [b] and would not be used to convey the
latter. In both [ii] and [iii], [b] entails [a], but the difference in meaning between [iiia]
and [iiib] is very much greater than that between [iia] and [iib]. As we have seen, the
only difference between [iia] and [iib] is that [iia] allows for the scenario in which I am
indifferent as to whether I’m included or not. But [iiia] allows for a much greater range
of possibilities than [iiib]: it could be, for example, that I want to be included, but am just
not insisting on it. It is only in the medium strength case – where there is relatively little
difference in meaning between matrix and subordinate negation – that the implicature
applies.
§5 Increased specificity of negation 841
The same holds for advice. Compare:

[5] i a. He didn’t allow me to go. b. He allowed me not to go. [weak]
ii a. He didn’t advise me to go. b. He advised me not to go. [medium]
iii a. He didn’t order me to go. b. He ordered me not to go. [strong]
Advise has medium strength in contrast to weak allow and strong order, and the impli-
cature is found only in case [ii]. Again, [ia] is much more informative than [ib] (which
it entails), and [iiib] is much more informative than [iiia] and could not be conveyed
by it. But the difference between [iia] and [iib] is much less. They don’t have the same
meaning, of course, since [iia] is consistent with his not having offered any advice at all.
If we assume that he did give some advice as to whether I should go, however, then [iib]
can be inferred from [iia]. Note that the implicature is more likely to be present when
advise is used performatively (Ch. 10, §3.1), as in I don’t advise you to go.
Likely expresses medium strength epistemic modality. Compare this time:
[6] i a. It isn’t possible that he’s alive. b. It is possible that he isn’t alive. [weak]
ii a. It isn’t likely that he’s alive. b. It is likely that he isn’t alive. [medium]
iii a. It isn’t certain that he’s alive. b. It’s certain that he isn’t alive. [strong]
As before, the implicature is found in the medium strength case [ii], but not in [i/iii]. In
[i], [a] entails [b] and is much more informative. In [iii], [b] is much more informative
than [a] and cannot be pragmatically inferred from it. But in [ii] there is relatively little
difference between [a] and [b]. Like many in [1iia], likely is an imprecise term: the lower
bound for what probability qualifies as ‘likely’ is fuzzy, just as it is for what number
counts as ‘many’. Suppose for the sake of argument we take the lower bound to be 60%:
if there’s a 60% chance he’s alive then we’ll accept It’s likely that he’s alive as true. In this
case, [iia] puts the probability of his being alive at less than 60%, while [iib] puts it at
40% or less. The interpretation of [iib] is then more specific by cutting out the middle
ground in the range 40% to 60%. Again this is comparable with the interpretation of
negated multal quantification as paucal.22
Such verbs as think and believe are also often used to express medium strength epis-
temic modality, but differ in taking two arguments, with the subject associated with
the role of experiencer (the one making the epistemic judgement). They contrast with
strong know, but there is no syntactically comparable verb or adjective expressing weak
modality. Compare, then:
[7] i a. She doesn’t think he’s alive. b. She thinks he isn’t alive. [medium]
ii a. She doesn’t know he’s alive. b. She knows he isn’t alive. [strong]
As before, [b] is an implicature of [a] only in the medium strength case: the difference in
meaning between [iia] and [iib] is too great to be pragmatically ignored. Note, moreover,
that although we have not provided a contrast with weak modality here there is no doubt
that think belongs in the medium strength category. The weak case can be expressed in
other ways, such as She is quite open-minded as to whether he is alive or (more naturally,
but with negation) She has no idea whether or not he is alive.
22
There is one catenative that allows the implicature which is semantically closer to multal often than to the
modal items in [3], namely tend : compare They don’t tend to read the fine print and They tend not to read the
fine print.
The items in the perception category [3v] are very similar, and although there are no
simple contrasts with strong or weak modality, it is clear that they belong in the medium
category. As before, [a] implicates [b] in the pair:
[8] a. He doesn’t seem to understand. b. He seems not to understand.
Conventionalisation of specificity increase

Not every item expressing medium strength modality permits the increased specificity
implicature that we have been illustrating. Lexemes meeting the general semantic con-
dition become associated with it on a piecemeal basis.
For example, while the implicature is found with likely, as illustrated in [6ii], it is
very questionable with probable: It’s not probable that he’s alive is not a natural way of
conveying “It is probable that he’s not alive”. Similarly, I don’t recommend that you tell
them conveys “I recommend that you not tell them” much more readily than I don’t
suggest you tell them conveys “I suggest you don’t tell them”.23 There are, moreover,
dialect differences with respect to some items, as indicated by the % annotation. Thus
for some speakers, but not others, I don’t guess there’s anybody home can convey “I guess
there isn’t anybody home”. At the same time, there are differences with respect to how
readily the construction with subordinate negation is used. Thus He seems not to have
understood is perfectly natural, whereas I want to not go is highly unusual, and normally
sharply disfavoured relative to I don’t want to go.
A clear indication of the importance of conventionalisation in this area is provided
by cross-linguistic differences in the items which allow the implicature. In English the
verb hope clearly belongs in the medium strength category, like want, but its nega-
tion is not pragmatically interpreted as applying to the subordinate clause. Thus I
don’t hope you’re late does not implicate I hope you’re not late. In German, however,
the corresponding verb, hoffen, does permit the implicature, behaving like the English
verbs in [3].
No subordinate negation implicature with modal auxiliaries

It will be noted that (leaving aside the had of the idiom had better) there are no modal
auxiliaries among the verbs in [3]. At first glance the following might appear to behave
in the same way:
[9] i You mustn’t tell anyone. [strong]
ii You shouldn’t take the job. [medium]
In both of these the negation is located syntactically in the matrix clause but applies
semantically to the subordinate clause: in this respect they are like the examples we have
been considering. There is, however, an important difference. The semantic association
of the negation with the subordinate clause is this time not a matter of implicature but of
sentence meaning. We noted in Ch. 3, §9.10, that a syntactically negative modal may or
23
Note also the difference between the epistemic and deontic senses of expect. In the epistemic sense (roughly
“think likely”) the implicature generally goes through, but in the deontic sense (“think x should”) its appli-
cation is more restricted. Consider, for example, We don’t expect them to pay more than $100. In the epistemic
sense this implicates “We expect that they won’t pay more than $100”. But in the deontic sense (roughly
“We don’t regard them as having an obligation to pay more than $100”) there is no subordinate negation
implicature: it doesn’t convey “We regard them as having an obligation not to pay more than $100”.
§6 Multiple negation 843
may not fall within the semantic scope of the negation. The examples in [9] have what we
call ‘internal negation’: the negation applies semantically to the non-finite complement
of the modal, not to the modal itself. Thus [i] imposes or reports an obligation not to tell
anyone. To negate the modality we use need: You needn’t tell anyone. Note, moreover, that
must expresses strong modality, whereas we have seen that the implicature only applies
in the case of medium strength modality.
Should, on the other hand, does belong in the medium strength category, but the
subordinate negation interpretation, approximately “The right thing for you to do is
to not take the job”, is still the meaning proper, not an implicature. Example [9ii]
differs from the examples with a subordinate negation implicature in that it does
not allow the less specific interpretation in which the negation applies semantically
to the matrix: it doesn’t mean “The right thing for you to do isn’t to take the job”.
This is evident from the fact that one can’t say: # You shouldn’t take the job and you
shouldn’t not take it either: it doesn’t matter whether you take it or not. The impli-
cature arises only in cases where there is a syntactic contrast between matrix and
subordinate negation like that found between [8a] and [8b], and the other pairs
discussed.
The can’t seem to construction

[10] a. I can’t seem to get it right. b. I seem not to be able to get it right.
The meaning of [a] is the same as that of [b] – or of It seems that I can’t get it right. Syn-
tactically, seem falls within the complement of the negated modal auxiliary, but seman-
tically it is outside the scope of the negative. This example differs from those considered
above, however, in that modal can also belongs semantically within the complement of
seem, rather than the other way round, as one would expect from the syntax.
This is another case where the interpretation is a matter of meaning proper, not
implicature: can’t seem (as used here) is simply an idiom meaning “seem unable”. The
essential parts of the idiom are can, negation, and seem; note, for example, that seem
is not here replaceable by appear. The mismatch between syntax and semantics here is
probably related to the fact that can has no plain form, so we don’t have ∗I seem not to
can get it right.
6 Multiple negation
When a clause contains two or more negative elements we need to distinguish cases
where they express separate semantic negations from those where only one semantic
negation is involved:
[1] i I didn’t say I didn’t want it. [two semantic negations]
ii He consulted neither his wife nor his parents. [one semantic negation]
In [i] we clearly have two semantic negations, one of want and one of say. In [ii], however,
there is just one semantic negation: it has scope over the coordination and is syntactically
expressed twice, once in each coordinate. The negation in [ii] is non-verbal, and the fact
that there is only one semantic negation is evident from the version with verbal negation,
which is semantically equivalent: He didn’t consult either his wife or his parents. Neither
and nor in [ii] are markers of coordination, and exhibit what can be regarded as negative
concord, or agreement.
Constructions with semantic negations in separate clauses, as in [1i] or I didn’t promise
[not to tell them] are unremarkable and do not need to be considered further (for the
special case where the negation markers are adjacent, as in You can’t not go, see §2.3.2
above). Nor do we need to say any more here about affixal negation. This is always
subclausal, and it can combine unproblematically with clausal negation, as in Their
behaviour was certainly not immoral or None of the problems seemed unimportant. In §6.1
we will look briefly at constructions with separate semantic negations in a single clause,
each of which could by itself mark clausal negation. Then in §6.2 we turn to concordial
and similar types of negation.
6.1 Multiple semantic negation within a single clause

(a) Constructions where the first negation has scope over existential quantification
[2] i None of them had no redeeming features.
ii No one, surely, has never experienced such temptation.
iii Never before had no one nominated for the position.
iv Neither investigator had no financial interest in the company.
v No one didn’t consider it a retrograde move.
By virtue of the relation between existential and universal quantification (see Ch. 5,
§5.1) these are all equivalent to positive clauses with universal quantification: All of
them had some redeeming features; Everyone, surely, has at some time experienced such
temptation; Before, someone had always nominated for the position; Both investigators had
some financial interest in the company; Everyone considered it a retrograde move. Note
that it is the first negative that is replaced by a universally quantified counterpart in the
paraphrase. Thus in [ii] no one precedes never, so we have everyone and at some time,
but in [iii] never precedes no one, so we have always and someone (even though in the
positive they do not occur in that order).24
The positive versions are of course easier to process and represent the default way
of expressing the meanings concerned. The more complex forms would thus typically
need some special motivation, such as contrast. In [2iii], for example, we are concerned
with an occasion on which no one had nominated for the position in question, and
[iii] contrasts this occasion with all previous ones of the relevant kind. Such contexts of
contrast will often lead to one or other of the negative markers being stressed.
As far as the syntax is concerned, only one of the negators can mark clausal negation.
The clauses in [2] are all negative and behave with respect to the tests for polarity just like
clauses with a single negative. Compare, for example, None of them had any redeeming
features, did they? and None of them had no redeeming features, did they?
24
The same kind of relationship holds between disjunctive and conjunctive coordination, so that Neither Kim
nor Pat had no financial interest in the company is equivalent to Both Kim and Pat had some financial interest
in the company.
§ 6.2 Negative concord and pleonastic negation 845
(b) Other types

[3] i [He didn’t say nothing:] he said it didn’t matter.
ii Not all of them made no mistakes.
iii Not many / Few people found nothing to criticise.
iv [We not only made no progress:] we actually moved backwards.
In [i] nothing follows the verbal negation, and this time the meaning involves existential
quantification: “He did say something”; examples of this kind are used to contradict a
negative assertion, in this case He said nothing. Example [ii] has negation of the universal
quantifier all and is accordingly equivalent to a positive with existential quantification:
Some of them made some mistakes. Example [iii], with negation of the multal quantifier
many or combination of negation with paucal quantification in few, has no positive
semantic equivalent of a comparable kind. But given a reasonably large set of people
under consideration it will tend to be pragmatically equivalent to Most people found
something to criticise.
One of the commonest cases has not only with scope over a negative, as in [3iv]. This
differs from the other types in that we can drop the not only without affecting the other
negative, giving We made no progress: we actually moved backwards.
6.2 Negative concord and pleonastic negation

We turn now to constructions where a single semantic negative is expressed more than
once.
(a) The standard variety

There are a number of constructions in Standard English in which the negation of a
clause is expressed at more than one point morphologically:
[4] i Their action was neither illegal nor immoral. [disjunctive coordination concord]
ii They aren’t here, I don’t think. [parenthetical concord]
iii Not in my car, you’re not. [negative retort]
iv I wouldn’t be surprised if it didn’t rain. [pleonastic subordinate negative]
Disjunctive coordination concord
When a negative has scope over disjunctive coordination, it may be expressed verbally, as
in Their action wasn’t (either) illegal or immoral, or non-verbally, in the coordination itself,
as in [4i]. In the latter case it is generally incorporated into all markers of coordination,
and we can therefore talk of concord or agreement in polarity. Occasionally, however,
we find or instead of nor: Their action was neither illegal or immoral (see Ch. 15, §2.4).
Parenthetical concord
A further minor case is seen in [4ii], where semantically there is a single negation, just as
there is in the non-parenthetical versions I don’t think they are here or I think they aren’t
here. In the parenthetical version the negation is expressed both in the anchor they aren’t
here and in the parenthetical I don’t think. The negation in the parenthetical, however,
is optional, for we can also have They aren’t here, I think. The matching negation in [4ii]
is comparable to the matching interrogatives in Are they here, do you know? (cf. Ch. 10,
§5.3).
Negative retort
Example [4iii] might be said by B in response to A’s saying I’m just driving into town.
B’s response is thus a reduced version of You’re not driving into town in my car. The
negated initial constituent represents new information, while the rest is discourse-old,
recoverable from A’s utterance. The effect is to emphatically reject a proposition or
proposal that is more specific than the one just uttered.
Pleonastic negation in subordinate clauses
Example [4iv] is ambiguous between a reading with two semantic negations (“It would
not come as a surprise to me if I were to learn that it didn’t rain”) and the one we
are concerned with here, where there is only one semantic negation (“I wouldn’t be
surprised if it rained”). In this second interpretation the negative in the subordinate
clause is pleonastic, an extra mark of something that has already been marked – in this
case, in the matrix clause. Other examples are:
[5] i No one can say what might not happen if there were another earthquake.
ii He is unable to predict how much of it may not turn out to be pure fabrication.
The range of constructions where this pleonastic not is found is very restricted. In [4iv]
the pleonastic not is in the protasis of a remote conditional where the apodosis has a
negated expression of surprise (cf. also I wouldn’t wonder if . . . ). In [5i–ii] it is in an
interrogative clause headed by modal may. In all three examples the subordinate clause
containing pleonastic not is strongly non-factual.
Negative concord in non-standard dialects

The clearest case of negative concord in English is found in non-standard dialects. This
book is of course a grammar of Standard English, but the negative concord phenomenon
is so widespread and salient that it deserves some mention here.
In many dialects, ranging from Cockney (spoken in the East End of London, England) to
African American Vernacular English (AAVE, formerly known as Black English Vernacular,
spoken in segregated African American communities in the USA), the absolute negators no,
no one, nothing, etc., are used in negative clauses where the standard dialect has the NPIs any,
anyone, anything, etc.:
[6] non-standard standard
i a. !He didn’t say nothin’. b. He didn’t say anything.
ii a. !You gonna spend your whole life b. Are you going to spend your whole life
[not trustin’ nobody]? [not trusting anybody]?
iii a. !Nobody here didn’t point no gun at b. Nobody here pointed any gun at
nobody. anybody.
Each clause contains just one semantic negation and in the standard versions it is marked by
a single negator. In the non-standard versions, however, it is marked by verbal negation and
also on all the existentially quantified elements in the clause. We accordingly have negative
concord between the verb and these elements.
Non-standard clauses with negative concord are characteristically homonymous with
standard dialect clauses containing multiple semantic negation, such as [2v] and [3i] above.
Standard He didn’t say nothing means “He did say something (it’s not true that he said
nothing”). The bracketed clause of [6iia] could likewise be used in the standard dialect
with the meaning “not being in a state of refusing to trust anyone”. In principle [6iiia]
could be used in Standard English to express a meaning containing four semantic negations,
§7 Positive and negative polarity in responses and anaphora 847
though in practice of course it would be far too complex to process. There are, however, some
constructions – such as imperative Don’t nobody move! – that cannot be used in the standard
dialect to express multiple semantic negation.25
There is an extremely widespread tendency among Standard English speakers to regard
dialects with negative concord as ‘illogical’ and ‘inferior’. It is argued that by a rule of logic
two negatives cancel each other out to make a positive. Thus just as It isn’t the case that she
didn’t move (or She didn’t not move) is equivalent to She moved, it is argued that He didn’t say
nothing is a double negative that can only mean “He said something” and hence should not
be used to express the opposite of that meaning. But such an argument is completely invalid.
The rule of logic 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. And as far as the [a] examples of [6] are concerned, there is only a single semantic
negation, so the rule of logic doesn’t apply: it is completely irrelevant.26 The pattern in the
non-standard dialect is similar to the one found in the standard dialect of Italian, French,
Spanish, Polish, Russian, and many other languages. For example, Italian non means “not”
and nessuno means “nobody”, but the meaning of Non ti credo nessuno is “Nobody believes
you”, not “Nobody doesn’t believe you”. Here again, then, we have two negative words marking
a single semantic negation, just as we do in the non-standard English dialects. There is no
more reason to condemn the latter as illogical than there is to condemn Italian, French, and
so on. The difference between the [a] and [b] versions of [6] is a matter of grammar, not
logic, and neither set can be regarded as intrinsically superior to the other.
Despite its non-standard character every experienced user of English needs to be passively
acquainted with the negative concord construction in order to be able to understand English
in such ordinary contexts as film soundtracks, TV dramas, popular songs, and many everyday
conversations. Those who claim that negative concord is evidence of ignorance and illiteracy
are wrong; it is a regular and widespread feature of non-standard dialects of English across
the world. Someone who thinks the song title I can’t get no satisfaction means “It is impossible
for me to lack satisfaction” does not know English.27
7 Positive and negative polarity in responses and anaphora
7.1 Answers to polar questions and comparable responses

Yes and no answers
Yes and no serve as markers of positive and negative polarity in answers to ques-
tions. They may stand alone, or combine with a clause that expresses the answer more
25
The same applies to the construction without + no (corresponding to standard without + any), as in !Give
me a large cheeseburger without no onions. And also to the construction with negative concord between the
verb and an approximate negator of degree, as in !I can’t hardly see (corresponding to standard I can hardly
see).
26
Given that the term ‘double negative’ is strongly associated with the semantic rule whereby two negatives
do cancel each other out, it is an unsatisfactory term for the negative concord construction. It is in any case
inappropriate because there is no limitation to two, as we have seen.
27
Negative concord was common in Old English and became virtually obligatory in the Middle English period.
Its decline in the standard written language in the early Modern period may have had much to do with a
nascent prescriptive tradition and its conscious comparison of English with Latin. In the nineteenth century
negative concord re-emerged as a literary mark of non-standard usage, the gap in the historical record almost
certainly concealing a continuous but largely unrecorded tradition in many spoken dialects.
explicitly:

[7] i a. A: Is this car yours? B:
Yes (it is). No (it isn’t).
b. A: Isn’t this car yours? B:

ii a. A: He has gone, hasn’t he? B:
Yes (he has). No (he hasn’t).
b. A: He hasn’t gone, has he? B:
The choice between yes and no depends simply on the polarity of the answer – not, for
example, on agreement vs disagreement with what may be suggested by the question.
Polar questions, especially negative ones, may be biased, indicating the questioner’s
predisposition to think that one or other answer is the right one, but that has no bearing
on the choice between yes and no. In [ib], for example, the appropriate response is yes
if the car is B’s and no if it isn’t, irrespective of what A appears to expect is the case.
Similarly with answers to tag questions, as in [ii].
∗
Yes it isn’t and ∗No it is are thus ungrammatical as single clauses. In Yes it is and No
it isn’t, the yes and no can be regarded as a special type of adjunct, a polarity adjunct,
which agrees in polarity with the clause – a further case of polarity concord in English.
The adjunct can also be placed at the end of the clause, with prosodic detachment: It is,
yes and It isn’t, no. A response to [7i] with the form No, it’s Kim’s would not of course
violate the polarity concord rule, because here we have not a single clause but a sequence
of two, just as we do in No it’s not mine, it’s Kim’s.
One respect in which the agreement vs disagreement factor is relevant concerns the
choice between single-word and expanded responses. Suppose you ask Didn’t you post
the letter after all, then?, indicating that you think I didn’t. If in fact I did post it, I would
normally say Yes I did, not just Yes.
Responses to other kinds of speech act

Yes and no are used in response to statements in a similar but not identical way:
[8] i A: She did very well. B: Yes (she did ). No she didn’t.
ii A: She didn’t do very well. B: Yes she did. No (she didn’t).
In [i] the disagreeing negative response would not normally be reduced to No, and in
[ii] the disagreeing positive answer could not be reduced to yes. It is in fact here possible
to say yes to express agreement with the negative statement: “Yes, you’re right”.
Following directives, yes and no can be used to express intention to comply with a
positive and negative directive. No is also used to indicate refusal to comply with a positive
one, but yes is not an idiomatic way of refusing to comply with a negative directive:
[9] i A: Remember to lock up. B: Yes (I will ).
ii A: Don’t forget to lock up. B: No (I won’t).
iii A: Tell me who did it. B: No (I won’t).
iv A: Don’t tell them I did it. B: ?Yes I will.
In [iv] B would more likely say just I will (tell them), or words to that effect. With
directives expressed by certain kinds of interrogative clause (a type of indirect speech
act, in the sense of Ch. 10, §9.6.1), a response may reflect the literal question meaning
or the indirect directive meaning. Thus in response to Would you mind coming a little
earlier next week, I might respond No of course not, I’ll come around six (“No of course
I wouldn’t mind”), or Yes of course, I’ll come around six (“Yes of course I’ll come a
little earlier”).
§ 7.2 Anaphoric so and not 849
Idiomatic negative answers with not and no

There are a number of idiomatic phrases that express an emphatic negative response to
a question or other speech act. Some have not, others the determinative no:
[10] i Not for all the tea in China! Not likely! Not on a bet! (AmE)
Not in a million years! Not on your life! Not on your nelly! (BrE)
ii No fear! No chance! No way! 28
7.2 Anaphoric so and not

With predicates that take so as the anaphoric pro-form for a positive clausal complement,
not is used as a pro-form for a negative (cf. Ch. 17, §7.7.2). We illustrate with answers to
questions, but they occur more widely than this.
[11] question positive answer negative answer
I believe/think so. I believe/think not.
Are they reliable? I was told so. I was told not.
It seems so. It seems not.
Not (like so) is here functioning as complement and marks non-verbal negation: it is not
modifying the verb.
What accompanies the pro-form need not be a full matrix clause; a preposed AdvP
functioning as clause adjunct will serve as well:
Most definitely so.

[12] Apparently so. Apparently not.
Is the city beautiful?
Most definitely not.
Other adjuncts such as PPs are permitted with not but not so:
∗
[13] On the whole so. On the whole not.
∗
Does it rain much? So in the winter. Not in the winter.
∗
Usually so this early. Usually not this early.
Not can also introduce anaphorically reduced clauses used in response to other types
of speech act or following a negative clause:
[14] i A: I think you should leave now. B: Not without my money.
ii I won’t go, not even if they beg me.
iii There aren’t many wild rhinoceroses left, not in Africa or in Asia.
Not is here understood respectively as “I won’t leave”, “I won’t go”, and “there aren’t many
rhinoceroses left”.
28
No way can also be integrated into clause structure as an emphatic negator, as in No way is that a diamond!
The original manner meaning has here been bleached away, so that we understand “That is emphatically not
a diamond”.

Negation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Negation

Uploaded by

Copyright:

Available Formats

9

1.1 Tests for clause polarity

Clause continuations with not even

Reversed polarity tags

Subject–auxiliary inversion with prenuclear constituents

1.2 An overview of negation types

[4] analytic synthetic

(a) Verbal vs non-verbal negation

(b) Analytic vs synthetic negation

We treat examples like He had no money as cases of synthetic negation because no

(c) Clausal vs subclausal negation

Clausal negation is a matter of syntax, while affixal negation is purely morphological.

(d) Ordinary vs metalinguistic negation.

1.3 Scope and focus of negation

1.3.1 The concept ‘having scope over’

Elements whose meaning is not truth-conditional

Here [b] can be analysed into three components of meaning:

Semantic and syntactic identification of scope

(a) Relative scope of negation and adjuncts in clause structure

(b) Relative scope of negation and quantifiers

Negation focus, falsity conditions, and prosody

Narrower and broader negation focus

Positive implicatures resulting from the choice of focus

Focus of negation as a special case of informational focus

2.1 Primary verbal negation

The choice between analytic and synthetic primary negation

Grammatical restriction on synthetic negatives: inadmissible in inverted conditionals

Position of negator in clauses with subject–auxiliary inversion

Subclausal primary negation

2.2 Imperative negation

Contrasts between imperatives and non-imperatives

(b) Order of subject and verb

(d) Restrictions on analytic negation

2.3 Secondary verbal negation

2.3.1 Formal marking of secondary negation

2.3.2 Secondary negation with not following an auxiliary verb

the following verb:

(a) Not immediately following a primary form of an auxiliary

(b) Not syntactically separated from an auxiliary

3.1 Not as a marker of non-verbal negation

(a) The not all type

(b) Not one

(c) Not two, etc.: “less than’’

(d) Not a little, not a few

giving “a fairly large amount/number” (compare the use of no in no small achievement in

(e) Not even, not only

(f) Not very, not quite, etc.

(g) Not unattractive: not with affixally negated adjectives

(h) Not unnaturally: negation of adverbs

(i) Not with PPs

(j) Not in verbless clauses

(k) Not + that clause

(l) Not in coordination

Unintegrated final not

ii Obviously the government is going to tell us the whole truth . . . not!

This construction is found mainly in younger-generation speech (popularised and perhaps

3.2 The synthetic absolute negators

3.2.1 Clausal negation

Alternation with verbal negation constructions

3.2.2 Subclausal negation

(a) Negative NP as complement of preposition

(b) No + mean, small, etc.

(c) Semantically clausal NP

3.3 The approximate negators