Negative Attraction and Negative Concord in English Grammar
Author(s): William Labov
Source: Language , Dec., 1972, Vol. 48, No. 4 (Dec., 1972), pp. 773-818
Published by: Linguistic Society of America
Stable URL: https://www.jstor.org/stable/411989
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to
Language
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
IN ENGLISH GRAMMAR
WILLIAM LABOV
University of Pennsylvania
The attraction of the negative to subject any is an invariant rule for all dialects
of English, and appears to respond to the distributive and hypothetical features
of this indeterminate. Negative concord, on the other hand, is a variable rule
which distributes the negative rightward in response to affective factors. It is
progressively extended from one dialect to another in an implicational series
which initiates negation in new positions as the rule approaches obligatory
status in the original environments.'
1. A SEMANTIC CONTRADICTION. The following sentence poses a difficult problem
for linguistic analysis, since it has two directly opposite meanings, depending
on who hears it:
(1) It ain't no cat can't get in no coop.
This is a sentence of the Black English (BE) vernacular spoken by Speedy J.,
the fifteen-year-old leader of the Cobras. It was recorded in a discussion of pigeon
flying during a group session in the course of recent studies in South Central
Harlem.2 For speakers of other dialects, this has the meaning of Standard English
(SE)
(2) 'There isn't any cat that cannot get into any [pigeon] coop.'
But this was not the meaning intended by Speedy. This will be clear if we place
ex. 1 in a larger context:
(3) CR: Do cats ... ever get in your cage?
Speedy: They never got in one of mine.
Junior: No; they scree-unless they got th-one of them JIVE coops!
[general laughter]
Speedy: [chuckle] It ain't no cat can't get in no coop.
Speedy's meaning is plainly (4), the direct opposite of (2):
(4) 'There isn't any cat that can get into any [pigeon] coop.'
Whatever the historical origins of BE may be, it is now a dialect of English.
In case after case, we find that striking differences from other dialects may be
traced to minimal adjustments in English phonological and grammatical rules
1 A preliminary version of this paper was given at the 1968 winter meeting of the Linguistic
Society of America, New York City, based upon work supported by the Office of
Education as Cooperative Research Project 3288. The present form of the analysis reflects
original contributions of Paul Cohen, as well as discussions with Edward Klima, Charles
Fillmore, Bruce Fraser, C.-J. Bailey, and Harris Savin, whose assistance is gratefully
acknowledged. The analysis of the semantic features of any in ?6 is particularly indebted to
recent work by Savin, whose views provide the point of departure for the logical characterization
of the quantifiers given there.
2 For an account of the field methods and the social background of the various groups
studied, see Labov et al. 1968. Most of the data on BE cited in the present paper is the product
of the field work of John Lewis and Clarence Robins.
773
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
(Labov 1969; Labov et al. 1968). There are differences in the selection of redundant
elements, different choice of dummy elements, and a few differences in
delicacy of verbal aspect.3 But on the face of it, ex. 1 poses a much more dramatic
difference which seems quite unrelated to aspect, emphasis, or style. For one
surface structure, we have two contradictory interpretations which must correspond
to the deep-structure difference.
This is not an isolated case. As we will see (?4), it is one of a half dozen examples
where the BE rules of negative concord produce forms that are interpreted
with exactly the opposite meaning by white listeners. The immediate problem
for the linguist-or for speakers of other dialects who would like to understand
BE-is to discover the nature of the rule which produces such an effect. What are
the conditions under which BE speakers can duplicate a negative in a following
clause without changing the meaning of the sentence? What process of change
could have caused a sentence which means X in dialect A to mean not-X in
dialect B? And how do speakers of A and B come to understand each other if
this is the case? The problem has even greater interest for our understanding of
linguistic evolution, since we do not yet understand the ways in which dialects
in close contact diverge to the point of unintelligibility and become distinct
languages (Labov 1971a).4
It appears that the problem of dialect differences in negative concord has not
been properly appraised. The naive schoolteacher's view is that the non-standard
dialects simply have too many negatives. Historically-minded linguists and
dialectologists point out that multiple negation is the traditional pattern, and
that our standard form is a rule imposed on English by grammarians in the 18th
century. More recently, students of syntax have pointed out that standard any
in He doesn't know anything already reflects the presence of a negative (or other
affective element) commanding it, so that He doesn't know nothing simply does
with the second no what SE does with any. Thus dialect differences are summarized
by Stockwell et al. (1968:272): 'The question of grammaticality for
double negation is complicated by the existence of a substandard dialect which,
like Chaucerian English, converts all some's directly into no's in negative sentences,
rather than leaving all but one of them as any's.'
As we will see, there are many dialects of English which convert some any's
into no's, differing from SE in a variety of ways, but there is no dialect which
converts all any's in negative sentences into no's. In other words, dialect differences
cannot be encapsulated by a some-no suppletion rule opposed to a someany
rule. In the course of our analysis of negative transfer rules, we will encounter
8 It is the invariant be registering habitual aspect which has drawn most attention in BE;
the preterit and future tenses are strongly represented, and the progressive and past perfect
as well. However, the present perfect shows many peculiarities in BE which have not
yet been well charted.
4 In some ways, BE is converging with other dialects of English, and reflects a Creole
origin with structures more different from English than we now observe. But so far as the
rules of negative concord and negative attraction are concerned, we are looking at the further
development of traditional, well-established English rules with no reflection in Creole
structures. We are therefore examining a divergence of dialect rules responsible for such a
phenomenon as sentence 1.
774
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
other evidence that dialect differences in the handling of negation ar
tensive than these authors suggest.
The problems posed by real language phenomena can best be hand
three-pronged attack. In this and other studies, we combine the abstra
of our intuitive data with naturalistic observation of language in use
plement this with experimental tests of well-defined variables. This
attempt to solve the semantic and grammatical problems posed by se
using several kinds of data to converge upon a single solution. ??2-4 w
the basic linguistic and dialectal data on three kinds of negative trans
tive attraction to subject any, negative postposing to indeterminates,
tive concord. In ?5 I will consider the possibility of combining these
into a single transformation. From a purely formal point of view this
possible. But when deeper questions are asked about what these transf
accomplish for speakers of the language, it appears that we are dealing
radically different kinds of grammatical operations. The strongly o
character of negative attraction provides a puzzle in itself, and in ?6 I
an analysis of the semantics of some, any, each, and all, in order to u
this categorical rule. In ?7, the variable nature of negative concord w
our attention, and it will appear that the contrast between categ
variable rules lies at the heart of our problem. ?8 integrates these fin
an over-all view of negative transfer rules; and from this over-all pa
answer to the problem raised by sentence 1 will be found.
2.1. NEGATIVE ATTRACTION. We must begin with the analysis of t
It ain't no cat, since this contains the semantic features which govern
formational rules operating in 1. It is generally agreed that the no
represents an underlying indefinite any combined with a negative which
attracted to it from elsewhere in the sentence. To understand the movement of
negatives in such sentences, we must deal with the whole range of rules which
govern negative attraction in English. These include the obligatory attraction of
the negative to subject indefinites(5)
*Anybody doesn't go -, Nobody goes;
the optional postposing of the negative to object indefinites(6)
He doesn't like anything . He likes nothing;
non-standard negative concord(7)
He doesn't like anything ' He don't like nothing;
and various forms of negative inversion(8)
He didn't ever go - Never did he go;
(9) Nobody saw it - Didn't nobody see it.
These rules all involve the attraction of NEGATIVE to the 'indeterminates' any,
ever, and either. This paper will be confined principally to these elements, without
taking up the problems associated with adverbs, prepositions, and conjunctions
that contain the negative feature.6 The label 'indeterminate' was first applied
6 Other indefinites and quantifiers are also involved in these rules; we will necessarily
consider some, and will also touch briefly on much and many, which lie on the periphery of
the rules being studied.
775
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
by Klima 1964 to distinguish any, ever, and either from other indefinites like some,
primarily on the basis of their co-occurrence with negative and question features.
The analysis of the quantifier no in no cat into NEG + any is a fundamental
given for our investigation. Klima assembles a great deal of evidence on this
point, which can hardly be questioned if one gives any weight to relatedness
between sentences. Typical is the alternation between active and passive forms
of sentences containing two indeterminates and one negative:
(10) a. No cook spoiled any broth.
b. *Any broth was spoiled by no cook.
c. No broth was spoiled by any cook.
There is no reason to think that most speakers would be aware of the underlying
any in no, but evidence from linguistic pathology indicates that a deep-seated
knowledge of this fact may in fact be present in most speakers. There is a particular
schizophrenic verbal pattern of reversing positive and negative in a
consistent and predictable way. A patient described by Laffal 1965 was exceptionally
gifted in the use of this rule:
(11) Doctor: Can you tell me now who invented the airplane?
Patient: I do know.
Doctor: You do know.
Patient: Yes, I know...
Doctor: What you mean to say is that you don't know.
Patient: I do know. If I don't know, I, I, I, I wouldn't be able to tell
you.
Doctor: You're not able to tell me, though, are you?
Patient: Yes I am, for I do know.
The patient could of course have reversed meaning simply by adding a negative
to every sentence. But he operated instead at a more abstract level, reversing the
negative modality of each underlying sentence by an alpha rule:
(12) [aneg] -, [-aneg]6
The effect is to remove any realization of an underlying negative. When the same
patient was performing a word-association test, he responded
(13) Home ... there's any place like
-demonstrating an unconscious knowledge of the NEG + any analysis of no
through the production of this totally ungrammatical sentence.
The negative is, then, normally attracted to the indeterminates any and ever,
realized as no and never. This attraction is obligatory for subject indeterminates,
but optional with others, so that we have the alternation:
(14) a. He didn't know anything about anybody.
b. He knew nothing about anybody.
But no indeterminate can be skipped:
(15) a. *He knew anything about nobody.
b. *He ever knew nothing about anybody.
6 This is one of the phenomena which justify the use of the feature notation [+neg],
implying the existence of a [-neg]. This patient's behavior could not easily be formalized
without the use of [-neg].
776
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
The English NEGATTRAC rule can be summed up informally by sayin
SE, the negative is attracted to the first indeterminate, obligatorily if it
ject. The simplest form of transformation that will accomplish this,
Fillmore 1967, operates upon the negative in its sentence-initial p
modality or type marker.7 Our first approximation to this transformatio
then be 16. (The category INDET stands for a set of features to be de
?6; the left bracket ([) stands for 'contains', and -[ for 'does not cont
(16) NEGATTRAC I
X - [+neg]- Y- INDET
1 2 3 4 - 1 3 2+4
Conditions: (a) Obligatory if
(b) 3 -[ INDET
Condition (b) insures that 4 is the fir
a first approximation to the statem
does NOT apply, then the negative wi
usual rule.
Even when 3 is null, and we are dealing with a subject indeterminate, there are
still many conditions under which NEGATTRAC is not obligatory. We will want to
examine these marginal cases in order to come to grips with the central phenomenon
of the categorical nature of NEGATTRAC. For when NEGATTRAC is
obligatory, it is one of the most mysteriously compelling obligations of all. On
the face of it, there seems to be no reason not to say
(17) *Anybody doesn't sit there any more.
But we can't say it, we don't say it, we won't say it; we reject it without hesitation
or reservation. For most listeners, 17 has a curiously ill-formed, fascinatingly
perverse character.
We get used to some ungrammatical sentences, and as we repeat them, we can
almost see how we might have said them ourselves. We can certainly imagine
someone saying
(18) *This bed was eaten potato chips in.8
If this is not already acceptable in some dialect, we would enjoy meeting the
7 Negative attraction is shown in 16 as a rightward movement from a leftmost abstract
constituent, but this view will not survive in the present analysis. Such transformations will
eventually be seen as leftward and rightward movements following negative placement in
the unmarked pre-verbal position. The ultimate location of constituents such as NEG or Q
in the underlying structure is of course currently in dispute. The analysis of negative attraction
and concord in the original version of this paper took these and other modalities
as features of a type marker which was a constituent of S in construction with NP and
PREDP. More recent views of NEG and Q as predicates of higher level sentences, reflected
in Carden 1970, are consistent with the present analysis. However, the analysis to be given
here is neutral on this issue, as it does not rely upon the formal presence of higher level S,
NP, and VP nodes which require lowering of quantifier or NEG.
8 This sentence cannot be reduced to the proper analysis for the passive transformation,
N1 - AUX - VB - NP2. There is considerable latitude in the analysis of various verb + particle
or verb + preposition combinations as VB; thus we have This bed was slept in, This wall
was stood beside. But the only VB + N + PREP combinations which can fit into this analysis
are idioms which have been lexicalized, like make fun of, take umbrage at. The acceptance of
18 involves accepting for the moment that eat potato chips in is such a VB.
777
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
man who had the imagination to say it. But we cannot imagine why anyone
would say 17, which becomes worse as we repeat it. We cannot even imagine
why we cannot imagine it; and if someone did say it, we are not at all anxious to
meet him.
2.2. CONDITIONS SUSPENDING THE OBLIGATORY CHARACTER OF NEGATTRAC.
Let us then consider some conditions where NEGATTRAC to subject indeterminates
is not obligatory, with the hope that the exceptions will aid in understanding the
rule. First we should note that a negative in the previous sentence has the effect
of making NEGATTRAC optional. Thus we can say, without hesitation,
(19) I don't say that anyone didn't go.
In place of the explicit negative of don't, we are also entitled to use the implicit
negatives in deny or doubt, or the presupposed negatives in regret, am sorry, or
apologize with the same effect on the rule. As we might suspect, there is a wide
range of acceptability of the incomplete or implicit negatives in suspending
NEGATTRAC. A feature [affect] combined with the negative seems to have greater
effect. In a series of test questions submitted to 52 student subjects,9 we observed
the following pattern of judgments:
GRAM- QUESTION- UNGRAMMATICAL
ABLE MATICAL
(20) a. I hated for anyone not to eat in 56 % 33 % 11 %
my mother's restaurant.
b. I expected anyone not to eat in 4 8 88
my mother's restaurant.
It is of course essential that the preceding negative be in the co
position to influence the NEGATTRAC rule; a preceding negative
clause has no such effect:
(21) *I didn't go and predicted anyone else wouldn't either.
The necessary condition is that the [+neg] in X of 16 must CO
determinate in the sense of TLangacker 1969-that is, must be a
which dominates the indeterminate and is not dominated by the in
We can then amend condition (a) of 16 to
(a) Obligatory if 3 = 0, 1 ~[ [+neg] commanding 4.
In addition to this condition, there are three other sets of e
which a subject indeterminate does not necessarily attract the n
quite different on the surface, but on deeper analysis they can be s
9 The tests were carried out at Columbia University by a group of st
mitted them to a range of non-linguistically trained subjects. They are
relevant but by no means conclusive evidence, as the subject populati
systematically or the tests administered in a uniform manner. Resear
grammaticality judgments is currently being carried out to confirm and
ings.
10 The conditions for removing the obligatory character of NEGATTRAC are quite similar
to those which govern the replacement of some by any cited in Stockwell et al. (262 ff.)
778
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
by a single feature that may be added to condition (a). Starting from
grammatical *Anyone can't go, we have:
(22) HYPOTHETICALS
a. If anyone can't go ...
b. When anyone can't go ...
c. Should anyone not be able to go ...
(23) RESTRICTED INDEFINITES
a. Anyone can't go who drinks.
b. Anyone who drinks can't go.
c. Anyone can't go unless he swears off drinking.
d. Anyone who is anyone won't go.
(24) EMBEDDINGS
a. For anyone not to go was a shame.
b. It's a shame for anyone not to go.
c. ?Anyone's not going was a shame.
The hypotheticals of 22 are all easily and clearly acceptable. They in
presence of a common feature which we may indicate provisionally
In 23, the subject indeterminate is limited by a restricted relative clau
tionally, relative clauses on any have also been understood as reduce
theticals (Karttunen 1971):
(23a') If there's anyone who drinks, he can't go.
If we follow this derivation, both 22 and 23 can be accounted for by
condition (a) of 16 the requirement that the type marker must not
[+hyp] if the rule is to be obligatory.
The nominalizations of 24 pose some difficult questions of syntactic and
tic analysis. First we may note that 24a-b are much more acceptable t
which is marginal. These reactions are supported in our set of test qu
GRAM- QUESTION- UNGRAMMATICAL
ABLE MATICAL
(25) a. Anyone doesn't eat in my 0% 4% 96%
mother's restaurant.
b. Anybody's not eating in my 31 55 14
mother's restaurant is a shame.
c. For anybody not to eat in my 73 27 0
mother's restaurant is a shame.
This is merely one of many findings which show that the P
nominalizers are not equivalents, but differ in subtle and
effects. One explanation for the effect of FOR-TO in elimin
character of NEGATTRAC is suggested by Klima (personal co
embedded sentences are almost always subjects of negative or
predicates. Since it has already been demonstrated that thi
important role in the syntax of indefinites (Klima 1964), th
be considered. However, the test results show surprisingly l
779
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
tween the positive 'O.K.' and the negative-affective 'shame':
GRAM- QUESTION- UNGRAMMATICAL
ABLE MATICAL
(26) For anybody not to eat in my 73% 27% 0%
mother's restaurant is a shame.
For anybody ... is O.K. 65 35 0
Whatever the affective character of shame may
smaller effect than that of embedding with the
ample, if the predicate O.K. is used with a TH
anyone, etc.), the corresponding percentages
in 25, switching to a POS-ING complementize
magnitude in judgments of grammaticality.
constraints inhibiting the obligatory force of N
which lists sentence types in order of increasin
The factors which favor the suspension of NEG
10 of Table 1: an affective implicit negative c
it is embedded in a subject with the FOR-TO
heavily weighted condition is the use of the FOR
7-10, as opposed to the POS-ING or THAT com
is this differentiation of FOR-TO and POS-ING to be understood? Our intuitive
response would be that the FOR-TO embedding suspends most completely any
actual predication, has the least 'predicating' effect of the three complementizers,
and is therefore closest to the IF clause, while THAT with its finite tense marker is
the furthest in the other direction. This general impression fits in with the efforts
of the Kiparskys (1968) to correlate the complementizers with factive and
non-factive predicators. While the data here are variable and at times contradictory,
there are enough clear cases to lend weight to the argument.
We find that non-factive predicators like want, prefer, believe etc. tend to take
the FOR-TO complementizer, while POS-ING regularly goes with such factives as
regret, approve, observe. Some verbs take either complementizer, and the difference
in assertiveness becomes obvious:
(27) a. I liked Albert to scratch his back.
b. I liked Albert's scratching his back.
The second appears to make it much clearer that Albert DID scratch his back. In
general, we can say that, whenever there is an option, the POS-ING complementizer
implies that the event took place, while FOR-TO makes no such claim.
There is a grammatical feature that 24 shares with 22 and 23, which is weaker
and more general than [+hyp]: it is a superordinate of [+hyp] which we will
designate [+fact]. The feature [+fact] is to be found in predicates containing
preterits, progressives, and perfects in the indicative; but sentences with subjunctives
or optatives, conditionals, or hypotheticals will show [-fact]. A sentence
embedded with a FOR-TO complementizer, without a tense marker, is
11 Extraposition plays a clear but minor role in promoting the acceptability of these sentences.
Sentence 10 may be compared with sentence 8 in this respect, and sentence 9 with
7. The gain in acceptability is approximately the same: a shift of ten to fifteen per cent out
of the 'questionable' column into the 'grammatical' column.
780
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
GBAM- QUESTION- UNGBAMMATICAL
ABLE MATICAL
1. Anyone doesn't eat in my mother's restaurant. 0% 4% 96%
2. I expected anybody not to eat in my mother's restaurant. 4 8 88
3. It's not true that anybody doesn't eat in my mother's res- 35 29 36
taurant.
4. Anybody's not eating in my mother's restaurant is a shame. 31 55 14
5. I hated anybody's not eating in my mother's restaurant. 56 33 11
6. It's O.K. that anybody doesn't eat in my mother's restaurant. 56 42 2
7. For anybody not to eat in my mother's restaurant is O.K. 65 35 0
8. For anybody not to eat in my mother's restaurant is a shame. 73 27 0
9. It's O.K. for anybody not to eat in my mother's restaurant. 79 17 4
10. It's a shame for anybody not to eat in my mother's restaurant. 85 15 0
(N = 52)
TABLE 1. Grammaticality judgments of fifty-two subjects on conditions for
obligatory negative attraction.
clearly non-factive and is to be assigned [-fact]. Thus we can say that NEGATTRAC
is not obligatory in 22, 23, or 24, since the X of 16 contains [-fact] commanding
the negative and the indeterminate.
It is not necessary for [-fact] to appear in the same clause as the negative and
indeterminate, only that it appear in a clause dominating them. The obligatory
character of NEGATTRAC is effectively cancelled in 22d:
(22) d. If John says that anyone shouldn't go, he's exceeding his au-
thority.
e. *Anyone shouldn't go.
This is because the [+hyp] if, implying the presence of [-fact] in the higher
clause, commands the negative and indeterminate.
The situation is not so clear with POS-ING complementizers. The 'questionable'
reactions to sentences 4 and 5 in Table 1 match our own question mark in 24c.
We are simply not sure if POS-ING is [-fact] or not. It is clear that POS-ING is
more assertive than FOR-TO, as in 27a-b. On the other hand, 27b asserts that
Albert did scratch his back much less clearly than 27c:
(27) b. I liked Albert's scratching his back.
c. I liked it that Albert scratched his back.
Rather than invent a new feature for POS-ING, it is best to say that, insofar as
POS-ING is perceived as containing [-fact], it suspends the obligatory character
of NEGATTRAC. That is, we have a match between two variable elements-the
factual character of POS-ING and its effect on NEGATTRAC.12
Finally, we must not fail to include one condition that can suspend the obligatory
character of NEGATTRAC: the presence of stress on any. The violations of
NEGATTRAC cited, as in 17, become more acceptable when we use stressed dny,
and perfectly grammatical if dny is preceded by just.
12 There are several examples in this discussion of features which are perceived to a degree
in specific lexical items, with uniformly vague reactions for many speakers. These
phenomena are more challenging to our system of discrete distinctive features than arguments
for n-ary as against binary features. Negative or non-factive features seem to be perceived
as present to a certain degree in some words or constructions.
781
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
(17') a. *Anybody can't eat there.
b. ?Anybody can't eat there.
c. Just anybody can't eat there.
We will refer to both the (b) and (c) possibilities as the presence of [+stress],
which relieves the strongly obligatory character of NEGATTRAC. Many linguists
would prefer to consider this stressed dny as a different item altogether (R.
Lakoff 1969, fn. 1). But such a strategy is appropriate only when we have no explanation
as to how stress added to any produces this effect. In ?6 such an explanation
will be submitted; thus we will proceed as if we were dealing with a
single any. Condition (a) of 16 must therefore indicate that any is unstressed if
NEGATTRAC is to be obligatory.
We will thus rewrite condition (a) as follows:
Condition (a): Obligatory if 3 = 0,
4 is [-stress], and
1 -[ [+neg] or [-fact] commanding 2 and
4.
This condition defines the profoundly ungrammatical core of NEGATTRAC violations
exemplified by sentences 1-2 of Table 1. A penumbra of questionable sentences
surround this core, as in 3-7, where NEGATTRAC is less strongly obligatory.
I have analysed this variable character as depending upon our variable perception
of a discrete feature, [-fact]. Though more might be said about this area,
it is sufficient here to note the minimal conditions under which the categorical
nature of NEGATTRAC is suspended, which has been done above. We can paraphrase
this condition in a general statement that also includes condition (b):
(28) Whenever an unstressed indeterminate any, without a negative feature
of its own, is not commanded by a negative or non-factive
feature, it may not be followed by a negative.
3. NEGATIVE POSTPOSING. Let us now consider in greater detail the optional
displacement of the negative to an indeterminate when Y in rule 16 is not zerothat
is, to the first object indeterminate after the verb. We will refer to this operation
as 'negative postposing'. No special conditions need be stated as long as
negative and indeterminate are within the same clause:
(29) a. I didn't find a proof of the theorem in any of these texts.
b. I found a proof of the theorem in none of these texts.
In most cases, the operation of negative postposing produces a marked form.
Both 14a and 29a seem to be neutral conversational forms, while 14b and 29b
are emphatic, perhaps formal or literary. There are some postposed expressions
which seem quite straightforward, even unmarked:
(30) a. That isn't anything new.
b. That's nothing new.
But these seem confined to the simplest and most common expressions involving
the contraction of the copula and other auxiliaries. Thus There's nothin' to it,
He's nowhere at all, You're nobody to talk all show the same unmarked character
as 30b. It is possible that these do not represent negative postposing at all, but
782
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
rather a different option at the level of contraction rules, yielding 's n
of isn't any. In any case, the vast majority of productive forms fo
tern of 29a-b, with negative postposing as the marked form.
When the indefinite is contained in a following clause, the situa
more complex. In some cases, the postposed negative allows both in
-that the negative originated in the earlier OR the later clause. Th
with 31, where (b) is ambiguous in this respect:
(31) a. I didn't tell John to paint any of these.
b. I told John to paint none of these.
As long as we are dealing with infinitival complements of the verb, it
such negative postposing can, with some strain on the imagination
clauses down from the original verb:
(32) a. He didn't order George to tell Arthur to ask Sam to d
like this.
b. He ordered George to tell Arthur to ask Sam to do not
this.
When the next clause is finite, negative postposing seems far more difficult. It
takes considerable imagination to give 33b the same reading as 33a:
(33) a. I didn't say that John painted any of these.
b. I said that John painted none of these.
When the indeterminate is in a relative clause, the situation becomes impossible
for negative postposing:
(34) a. I'm not going to sign a petition that any half-baked Stalinist wrote.
b. I'm going to sign a petition that no half-baked Stalinist wrote.
There seems to be no way that 34b could be derived from 34a; it is not clear that
34b means anything at all. We are dealing here with the inverse of Ross's Complex
NP Constraint (1967), since we are prevented from moving a negative into a
sentence dominated by a lexical head noun.
We thus have at least two additional constraints on Y in 16 beyond the fact
that it cannot contain an indefinite: the 'Inverse Complex NP Constraint' and
the 'Finite Complement Constraint'. Both are difficult to state formally; both
show exceptions and ragged edges. This is to be expected, because negative postposing
is an odd, irascible, cranky kind of transformation that has very shallow
roots in speaker intuitions.'3 To cover these two conditions, we can add to 16 the
following approximations to a formal statement:
Condition (c) ('Inverse Complex NP Constraint'): 4 cannot be dominated
by an S dominated by a lexical NP which does not
dominate 2.
Condition (d) ('Finite Complement Constraint'): 4 cannot be dominated
by an S adjoined to THAT and commanded by 2.
Both constraints could be examined in greater detail, but it should be obvious
13 We would expect negative postposing to have a weak intuitive base because it is so
much a part of the literary pattern learned as a superposed variety. This in turn reflects
its origin in the normative patterns imposed upon standard English grammar.
783
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
that the conditions which control NEGATTRAC when Y is NOT zero are very different
from those which operate when Y is zero, as discussed in ?2.1. Conditions
(b), (c) and (d), which control negative postposing, share no features in common
with condition (a), which governs NEGATTRAC to subject indeterminates.
4. NEGATIVE CONCORD. In various non-standard dialects of English, our formulation
of negative attraction must be extended to account for negative concord.
Instead of saying that the negative is attracted to the first indeterminate, we
might say for these dialects that the negative is attracted to indefinites generally.
Thus the non-standard equivalents of 29-34 are:
(29') I didn't find a proof of the theorem in none of these texts.
(30') That ain't nothin' new.
(31') I didn't tell John to paint none of these.
(32') He didn't order George to tell Arthur to ask Sam to do nothing
like this.
(33') I didn't say that John painted none of these.
(34') I'm not going to sign a petition that no half-baked Stalinist wrote.
Teachers and other opponents of non-standard dialects may argue that these
sentences reverse the meaning of 29-34. But this is mere rhetoric; any speaker of
English, no matter how refined, is familiar with the existence of negative concord
and realizes that 29'-34' intend the same meaning as 29-34. When an underlying
double negative is intended, speakers of non-standard dialects use the
same device as speakers of SE-heavy stress on both negatives:
(31") I DIDN'T tell John to paint NONE of these; I wanted to get some of
them painted at least.
(See ?8, ex. 145, for an example of this type of double negation in BE.)
The ordinary meaning of 29'-34' is therefore recognized by speakers of all dialects,
and these sentences do not produce the reversal of expected meaning observed
in ?1. The general non-standard rule which operates here can be written
as a simple pleonastic transformation, copying instead of chopping the negative:
(35) NEGCONCORD I
X - [+neg] - Y - INDET
1 2 3 4- 1 2 3 2+4
Conditions: (a) Obligatory if
4 is [-stress], and
1 -[ [+neg] or [-f
2 and 4.
Here we must repeat the same condition (a) that was worked out for rule 16,
since it applies to non-standard dialects as well as to SE. Condition (b) on 16,
which governs negative postposing, is missing. So too is the Finite Complement
Constraint (d), as indicated by 33'. None of the difficulty or strain in negative
postposing to remote infinite complements, as felt in 32-34, appears here-as indicated
in 32' and 33'. Even the Inverse Complex NP Constraint (c) does not
apply, as 34' shows.
It would be going too far to say that there are NO constraints on NEGCONCORD.
We find that we cannot transfer a negative to an indeterminate in a relative clause
784
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
if the indeterminate is a direct object. Thus 34 and 34' show NEGC
SUBJECT indeterminate in a relative clause, possible for non-standa
CORD but impossible for standard negative postposing. With OBJEC
nates in relative clauses, any negative transfer is prohibited:
(36) a. They wouldn't fire a man who could fix anything in a F
b. They would fire a man who could fix nothing in a Ford.
c. They wouldn't fire a man who could fix nothing in a Ford.
All three of these sentences have different meanings. We therefore
a condition (c) on NEGCONCORD as:
Condition (c): 4 cannot be directly dominated by a VP domin
S dominated by a lexical NP which does not dominate 2.
Except for this one condition, we can say that NEGCONCORD applies
regularly than negative postposing, and without the Finite Comp
straint and the whole series of marginal judgments involved in it.
construct such extreme examples as 37, with eleven pleonastic neg
surface structure all dependent upon a single negative in the deep
(37) I ain't [1] gonna sit here in no [2] chair and let no [3] crazy law
[4] tell me no [5] lies about no [6] law that no [7] judge t
smart-ass clerk to look up in no [9] book thatno [10] smart p
wrote or nothin' [11] like that nohow [12].
Here we see that the negative can be reduplicated in co6rdinate str
in subject indeterminates in relatives [7, 10], and also in oblique object
nates in relative clauses [9]. But copying on to a direct object indet
a relative clause seems to lie just outside the operation of this rule.
Is there an equivalent of condition (b) for NEGCONCORD? That is,
that Y cannot contain an indeterminate? If we can establish enou
in the conditions on 16 and 35, then it might seem practical to combin
a single transformation showing the over-all system of English negativ
If 35 applies to its own output, successively transferring the negative
then we could write such a condition (b). But that would require
determinate be skipped in the regular application of the rule. At th
tuitions cannot be relied upon, and we have to examine the evide
speakers of the non-standard dialects concerned actually say. Fro
Banks of North Carolina, we have:
(38) We don't ever see none of them guys. (Monnie 0., 35, Wa
And from a railroad foreman in Georgia:
(39) From then on, I didn't have any trouble at school, no mo
G., 60, E. Atlanta)
And from a Black speaker raised in the Lower South:
(40) Ain't nobody ever thought about pickin' up nothin'. (Wil
Florida)
Since it is possible to skip an indeterminate, then 35 can apply to any indeterminate
anywhere in the sentence, and there can be no constraint parallel to
condition (b) of 16.
785
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
4.1. NEGCONCORD TO PRE-VERBAL POSITION. In addition to the transfer of
negatives to indeterminates, we also have the possibility of negatives appearing
in the usual pre-verbal slot as well. There are many non-standard dialects which
do not allow this, and reject such sentences as Nobody don't know.l4 But this is
a regular construction in many non-standard dialects throughout the United
States. It is used in the white Southern dialect of the speaker quoted in 39:
(41) Nobody don't like a boss hardly.
We have not yet mapped all the dialects which permit transfer to the pre-verbal
position as against those which do not. Northern New Jersey seems to permit this,
while the white New York City vernacular definitely does not. We will use the
cover terms WNS1 and WNS2 to refer to the groups of white non-standard dialects
which respectively prohibit and permit negative transfer to pre-verbal
position.
Since white New Yorkers are members of the WNS, group, they contrast
sharply on this point with Black speakers in the same city. Negative transfer to
pre-verbal position is quite common in BE:1'
(42) Down here nobody don't know about no club. (William T., 25, Flor-
ida)
(43) None of our friends don't fight 'im. (Larry H., 15, Jets)
(44) Nobody don't know where it's at. (Douglas S., 17, Cobras)
(45) None of 'em can't fight. (Ray L., 14, Jets).
We can adjust our rule systems in one of two ways to account for this extension.
In 35' we can show VB as well as INDET as the locus of transfer:
(35) X - [+neg] - INDET
1 2 3 4 -I 1 2 3 2 + 4
The other possibility for dealing with pr
rule of negative placement put the negat
know. Since that is a simple categorical ru
will have to change the NEGATTRAC rule (16
dialects, and a copying rule for WNS2 an
have to be OPTIONAL, since NEGCONCORD
dialects. To insert this kind of variability
to its uniform, non-affective character. It is
verbal option a feature of the NEGCONCORD
affective character of that rule.
We have no choice in the matter when we observe that a dialect also exists
14 Bellugi (1967:143) shows that the children acquired a 'thick layer' of negation in the
later stages of their development. In his fourth year, Adam began to use NEGCONCORD from
subject indefinites to pre-verbal position, e.g. No one can't even find me; But nobody wasn't
gonna know it. Bellugi argues that this is an internal generalization, since it is not found in
any adult dialect. But as shown here, adult models are available for this pattern.
15 In the citations given, speakers are identified by their age and the area where they were
raised from 4 to 13 years, except for members of the peer groups in South Central Harlem
studied in Labov et al. These (all New Yorkers) are identified with the names of the peer
groups: Jets, Cobras, Thunderbirds, Aces, and Oscar Brothers.
786
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
which transfers the negative to pre-verbal position in a FOLLOWIN
almost all speakers of English, this is an impossibility. If a negative
a verb in a following clause, it is inevitably interpreted as referrin
deep-structure negative.
(46) None that I had, I never had any that I couldn't break yet
15, Sonora, Texas)
This sentence is rightly interpreted to mean that there is no hor
could NOT break. The fact that negatives in following clauses are
understood as independent is registered in our rule system by addi
condition that 2 and 4 must be clause mates-that is, there is no S
which does not dominate 4 and vice versa. But this constraint does
BE. We can see that ex. 1, It ain't no cat can't get in no coop, viol
negative in can't is copied from the first clause and has no independent
This is not an isolated instance. We have collected a number of
amples in our work with BE. A sales clerk in Harlem remarked one
(47) When it rained, nobody don't know it didn't.
That is, nobody knew that it rained when it DID, not when it didn
old man raised in the Bronx said:
(48) Back in them times, there ain't no kid around that a
even thinkin' about smokin' no reefers.
That is, there wasn't any kid who WAS even thinking about smo
And in one of the long epic poems of Black folklore, 'The fall', we
lowing description of the central character:'6
(49) Like a sex machine, she stood between
Raindrops, snow and hail;
She stood on hot bricks to lure her tricks
Come cyclone, blizzard or gale.
She tricked with the Frenchmen, torpedo men 'n' henchme
To her they were all the same;
She tricked with the Greeks, Arabs and freaks,
And breeds I can not name.
She tricked with the Jews, Apaches and Sioux,
She even tricked in the house of God,
For there wasn't a son of a gun who this whore couldn't shun,
That claimed to play a rod.
The meaning of the next to last line is plainly that 'there wasn't a son of a gun
who COULD shun this whore', and we have the same reversal of the expected meaning
as in 1, 47, and 48. We also recorded the following interchange between our
field worker John Lewis and Derek, one of the verbal leaders of the Cobras:
(50) JL: What about the subway strike?
Derek: Well, wasn't much I couldn't do.
16 The version of 'The fall' given here was recorded by John Lewis from Saladin, the most
skilled and knowledgeable of all masters of the oral literature that we have met. He learned
his version of 'The fall' in prison.
787
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
Here Derek plainly meant that there wasn't anything he COULD do. No indeterminates
are involved here; this is a case where the NEGCONCORD rule must transfer
the negative to the next clause (possibly governed by the quantifier muchsee
?7.3 below), and again there is no way for the regular rule of negative placement
to account for it.
It should be emphasized that such extrapositions of the negative to following
verb phrases are relatively rare. The normal construction in BE is NOT to apply
the NEGCONCORD rule in this environment:
(51) a. It wouldn't of been nothing I could do. (Larry H., 15, Jets)
b. Ain't nothin' you can do for 'em. (Roy M., 56, South Carolina)
Furthermore, speakers of BE have no difficulty in using the negative position in
a following clause for an underlying negative with independent meaning:
(52) I was writin' for fun 'cause I ain't do it till nothin' was happenin'.
(Florence, 13, West Philadelphia)
Here Florence is explaining that she didn't write with a red grease pencil on a
car until there was nothing happening in the neighborhood. Nevertheless, the
possibility of using NEGCONCORD with a following verb phrase is open to all
speakers of BE, who interpret 1 and 47-50 correctly and automatically.
There is no way to account for this possibility if we transfer the negative to preverbal
position by the usual cyclical rule of negative placement, since that rule
could not possibly transfer a negative to a following clause. We therefore conclude
that 35' is the correct expansion of NEGCONCORD; the condition that 2 and
4 are clause mates applies to WNS2 and not to BE.
We can sum up this view of the distribution of NEGCONCORD by an expanded
version of rule 35, following the same order of conditions used with NEGATTRAC
and negative postposing:
(53) NEGCONCORD II
X-[+neg] -Y- INDET
1 2 3 4 -- 1 2 3 2+4
Conditions: (a) Obligatory if
4 = INDET, [-stress]; an
1 [[ [+neg] or [-fact] co
(b) (c)
4 cannot be directly dominated by a VP in
dominated directly by an NP which does n
command 2.
(d) (e)
4 = VB in WNS2 and BE only; in WNS2, 2 and 4
must then be clause mates.
5. ONE NEGATIVE TRANSFER RULE? At this point we may feel compelled by
principles of economy, or by a belief in one over-all system of English grammar,
to write a single rule for NEGATTRAC, negative postposing, and NEGCONCORD,
788
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
combining 16 and 53. We can do this by making the pleonastic elem
(54) NEGTRANSFER
X-[+neg] - Y-INDE
1 2 3 4-- 1 <2>1 3 2+4
Conditions: (a) Obligatory if 3 =
4 = INDET, [-stress];
1 -[ [+neg] or [-fact
2 and 4.
(b) Y '[ INDET if < >1 is not realized.
(c) 4 cannot be (directly dominated by a VP)i in an
S dominated directly by an NP which does not
command 2.
(d) 4 cannot be dominated by an S adjoined to THAT
and commanded by 2 if < >1 is not realized.
(e) < > not in SE; 4 = VB in WNS2 and BE only,
in WNS2 only if 2 and 4 are clause mates.
This combined rule effects some economies, although certain conditions are
awkward to write. The pleonastic operation of concord is represented by an option
in angle brackets, with a subscripted cross-reference to the pre-verbal possibility.
This has an additional set of parentheses, indicating that it is optional
even if the subscripted option 1 is taken.
Conditions (b) and (d) are cumbersome to state because they are constraints
only on the standard rules, and we have to mark them formally as operating
only if option 1 is not taken.17 On the other hand, the subscripted insert on condition
(c) condenses the conditions for standard and non-standard rules with considerable
economy.
Condition (e) uses ad-hoc dialect labels. We could resolve these into a set of
features, beginning with [=tStd] and [d=BE]. We would then have to add a
feature [-NS] to differentiate WNSi and WNS2, although there is no match
with 'Northern States' vs. 'Southern States' in this distribution.'8 We would
then have:
(54') NEGTRANSFER
S -[+neg -X - jNDET }
1 2 3 4 - 1 <2>_std 3 2 + 4
Conditions: (a) Same as in 54.
(b) X -[ INDET / [+Std].
17 The angled brackets are used here in the conventio
reference within the rule, to indicate co-occurrence re
Constraint is also a variable constraint; if we were abl
want to show the rule operating more often with nonentirely
to them.
18 We would eventually want to replace the arbitrar
dialects located in space or time, filling out the paradig
789
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
(c) 4 cannot be <directly dominated by a VP>-std in
an S dominated directly by an NP which does
not command 2.
(d) 4 cannot be dominated by an S adjoined to THAT
and commanded by 2 / [+Std].
(e) If 4 = VB, 2 and 4 are clause mates / [-NS,
- BE].
This set of features gives us a more economical and less awkward representation
than 54. It is possible that such a limited set of features will be satisfactory, but
it is more likely that as our knowledge of NEGCONCORD grows, the number of
sub-classifications will increase rapidly. Rather than a set of conditions, we will
want to set up matrices of the kind to be developed in ?8.
One difficulty with 54' is that the registration of dialect options makes it awkward
to show variable constraints on the rule, and we will eventually want to do
this. A deeper understanding of NEGCONCORD must include the fact that the rule
applies more often in some environments than in others-e.g., more often when
4 = INDET than when 4 = VB.
The effort to produce a single rule schema can be rewarding in that it forces
us to define the precise nature of the differences between dialects. If it also
showed us interactions between conditions, it might actually explain why the
rule develops as it does. Those who believe in internal evaluation measures
would also believe that this condensation effort will help us select the correct
form of the grammar and even give us such explanations.l9 But without a direct
demonstration of such a claim, we must admit that 54' explains very little. In
fact, it obscures. A fundamental objection to 54' is that it combines two opposing
grammatical processes: NEGATTRAC on the one hand, and negative postposing
and NEGCONCORD on the other. NEGATTRAC is an obligatory process governed by
abstract, conceptual constraints. Negative postposing and NEGCONCORD are
optional, emphatic transformations. Since the argument of this paper will be
seen to revolve about the difference between these two types of rules, the major
point will be lost in such a global formulation.
The next two sections will examine the distinct semantic character of these
distinct types of grammatical rules. ?6 will explore the logical basis for NEGATTRAC,
which will allow us to unify and simplify the conditions which govern the
obligatory application of this rule. Instead of a rule which is obligatory in some
environments, we will derive a simpler statement of the underlying logical configuration
which is always prohibited in English and propose an explanation for
the prohibition. ?7 will then examine the affective character of NEGCONCORD, and
place this rule within a larger context of devices for reinforcing negation. These
findings will then be combined in ?8, which will attack the problem posed at the
outset: how can closely related dialects develop opposite interpretations of the
same surface structure?
'1 Lakoff 1971 notes (fn. 2) that the only explicit application of an internal evaluation
measure in syntax is Chomsky's analysis of the auxiliary in Aspects. In the light of Ross's
refutation of this, Lakoff argues that' there is not the slightest reason to believe that evaluation
measures play any role in the theory of grammar at all, much less in the innate biological
mechanisms of the child.'
790
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
6. THE NATURE OF NEGATTRAC. At the end of ?2, we arrived at a
principle (28) which summarized the conditions under which NEGAT
obligatory: 'Whenever an unstressed indeterminate any, without a n
feature of its own, is not commanded by a negative or non-factive f
may not be followed by a negative.' While this output condition will pr
correct results, it is both miscellaneous and unmotivated. It is not cl
stress, non-factives, and negatives should appear together as a list of e
to the obligatory character of NEGATTRAC; and there is nothing here whi
us to understand why violations of the obligatory core of the rule are
pretable, or why they produce an incurable confusion in listeners of e
and background.
The generality and force of NEGATTRAC to subject indefinites cannot be
stated. It is equally strong in all the English dialects we have studied.
have considered data from the abstract analysis of our own intuitions,
the observation of natural speech. We can also bring to bear the results of
ments with repetition tests which reveal the uniform character of the NEG
rule. These tests, carried out with groups of adolescent speakers of BE
that deep-seated rules of the vernacular influenced perception and p
more than had been realized.20 For a central core of vernacular speakers, B
were repeated verbatim, and certain SE forms were immediately transl
the corresponding BE forms. However, sentences which violated NEG
produced a different response: profound confusion. Here is the record
subjects' attempts to repeat a sentence that violated output constraint
(55) Anybody doesn't sit there anymore, do they?
Boot [1st trial]: Anybody-Hey, you goin' too fas'!
[2nd trial]: Any-I can't say it; I owe you a nickel.
David [1st trial]: Hunh?
[2nd trial]: That don't sound right!
Money [1st trial]: Anybody-eh-what is that?
[2nd trial]: Anybody ever sits there d-any more, do the
[3rd trial]: Anybody nev-ever sits thereBoot
and David respond with total confusion. Money tries to solve the
by removing the negative element; a fourth member of the group fol
same pattern, producing Anybody ever sits there anymore, do they?
Recent work with a wide range of English dialects shows that this c
response is not limited to BE speakers: violations of NEGATTRAC are
focus on, and impossible to interpret.
6.1. THE SEMANTIC FEATURES OF any IN POSITIVE SENTENCES. Resp
violations of NEGATTRAC suggest the presence of deep-seated logical in
which is not easily resolved by listeners trying to interpret such sentences
therefore be necessary to consider the general semantic features of any, fi
in positive sentences. This is an extraordinarily complex area with a lon
of resistance to analysis, and only some of the problems will be conside
20 See Labov et al., 1:3.9. Broader applications of such repetition tests have been
out on school populations, but violations of standard rules, as in 55, have not been
791
This content downloaded from
:ffff:ffff:ffff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
The discussion will be directed at uncovering just those central features of the
quantifiers which are responsible for the phenomena outlined above.
Our current approach to the study of language in context is based on the
strategy of a regular interplay between observations of natural behavior, the exploration
of intuitions, and experimental testing (Labov 1970, 1971c). The analysis
in this section starts from the empirical findings of the use of NEGATTRAC
and NEGCONCORD in various dialects, and adds data from my own intuitions and
those of other linguists, along with the evidence and arguments of scholars and
philosophers. Some of the intuitive data will rest on a firm foundation of general
agreement, using only well-known examples. But other sentence types may touch
on unexpected areas of variability in judgment, and may contain uncorrected
biases produced by the theoretical notions themselves.2' Recognizing the possibility
of such a bias, I present this analysis with the hope that it will illuminate
the earlier studies and lay the foundation for further empirical investigation.
Klima originally assigned the label 'indeterminate' to the class of any and
ever on the basis of the formal properties noted in ?2. His analysis of the behavior
of any has semantic implications, since he notes its co-occurrence with WHand
verbs such as wonder, which he marks with the general semantic feature
[+affect]. Nevertheless, Klima sees any in negative sentences as a formal suppletive
alternant of some, usually obligatory but occasionally optional with unanalysed
semantic consequences. Fillmore distinguishes two senses of some as
[=-specific]; he argues that it is the [-specific] which alternates regularly with
any. In this view, any remains an automatic suppletive alternant in negative
sentences which does not convey information of its own. Stockwell et al. (255-
69) review these various possibilities and point out difficulties with the [?specific]
hypothesis, but conclude that this is the most plausible account now available
for the matter. Accordingly, they represent the some-any suppletion transformation
with a structural analysis
(56) X - [+affect]-Y-Y - -det]-Z
and instructions to change [-indet] to [+indet]. This formal analysis of any
lacks the substantive semantic elements which would account for a large body of
linguistic facts, including the obligatory character of NEGATTRAC. The feature
[-indet] is simply a label; if 'indeterminate' had any significance at all, [-indet]
would be oddly redundant. R. Lakoff 1969 demonstrates the existence of meaningful
alternation between some and any. Typical are the oppositions:
(57) a. Who wants some beans?
b. Who wants any beans?
(58) a. Do you think those men want to do some work?
b. Do you think those men want to do any work?
Lakoff argues that some has a positive presupposition, any a negative one, and
that this must be entered in the formal description by some mechanism not yet
elaborated.22
21 Current empirical research on variation in intuitive judgments includes the re-examination
of many examples cited here to detect such individual or social differences.
22 She notes that there appear to be some 'purely syntactic constraints' on the positive
792
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
These are some of the many considerations which would lead us to
any embodies a set of semantic features distinct from some, in q
negatives as well as in positive declarative sentences. As a quantifie
limiter and designator of the extent of reference of the nominal being
Some is an indefinite quantifier, as opposed to definite quantifiers s
and is partitive as opposed to all and every. We can sum up these ch
in feature notation by saying that some is [+quantifier, -definite
or refer to it in logical terms as the existential quantifier which binds
propositions (3x)Fx.
Any is also [+quantifier, -definite]; but it differs from some in a
other semantic features-[+ distributive], [-partitive], and [-fa
consider these three features in positive contexts before going on to ex
interaction with the negative.
6.11. [+distributive]. Any is [+distributive], in that it considers
at a time. Any shares this property with each, as opposed to some
one person can do a given activity, the difference comes out strongly.
three bridge players, you can turn to a crowd at the other end of
ask
(59 Can any of you make a fourth at bridge?
But you cannot ask
(60) *Can some of you make a fourth at bridge?
By the operation of the same semantic feature, you would be invit
you asked
(61) Can some of you play bridge?
6.12. [-partitive]. Any is generally equivalent to every, each, and all in its
universal [-partitive] feature, as opposed to the existential [+partitive] character
of some, a, and other quantifiers. In terms of truth values, any and every
are equivalent in
(62) a. Any boy can run a mile.
b. Every boy can run a mile.
These statements both apply to the class of all boys, and to each and every boy;
for one to be true, the other must be true. There are also ambiguous occurrences
of any which make it seem that there is a [+partitive] any in hypothetical sentences.
The following pair cited by Fillmore (personal communication) illustrates
such ambiguity:
(63) a. If John can do it, anyone can do it. [-partitive]
b. If anyone can do it [t-partitive], John can do it.
In 63a, the conclusion is that 'everyone' can do it, and is plainly derogatory. But
63b has two readings. The [+partitive] one is complimentary to John: 'If there
use of any as in *Anyone left. The analysis given below shows this constraint as a regular
product of the semantic features of any and the preterit.
" In a recent discussion of logical properties of quantifiers, Keenan 1971 proposes three
characteristic properties: binding, relative scope, and predication. All the items discussed
here satisfy these requirements.
793
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
exists any one person who can do it, that person is John.' 24 The [-partitive]
meaning has the same derogatory interpretation as in 63a: 'If everyone can do
it, John can do it.' The ambiguity is then quite discrete, since the one meaning of
any is a knock, and the other is a boost.
This ambiguity of any is present in any hypothetical clause, but can be seen
most readily when we construct different consequences to the same antecedent.
(64) a. If John can sleep with any of these girls, then at least he'll have a
roof over his head. [+partitive]
b. If John can sleep with any of these girls, then he'll have a decision
problem. [-partitive]
One way to display the specific difference between these two uses of any is to
consider what we would have to do to test the validity of 64a-b. In 64a, the
consequence must be tested each time that a girl says yes and the antecedent
holds; if each girl who says yes has an apartment, then 64a is true. But for 64b,
no test of the consequence is relevant or necessary until the condition is seen to
hold for each and every individual: after all the girls said yes, we could then test
64b by seeing if John indeed had trouble deciding what to do.
This view of the matter is consistent with the analysis of the ambiguity of any
proposed by Savin 1971, in which 64a-b differ in the relative scope of the quantifier.
Thus we can represent 64a in the following logical configuration:
(65) (Vx) (John can sleep with x v John will have a roof over his head)
where the scope of the universal quantifier extends over both the antecedent and
the consequent. This is the characteristic property assigned to any by Reichenbach
(1947:106)-'that its scope is always the whole formula.' 25 But 64b shows
a different scope for any-in this case the more limited scope which Reichenbach
restricts to every:
(66) ((Vx) (John can sleep with x)) (D John will have a decision problem).
In this analysis, any is always a universal quantifier, or [-partitive] in our feature
notation, and the ambiguity is a structural rather than a substantive one.
The ambiguity of 63b would then be represented as
(63b') ((Vx) (x can do it)) D (John can do it).
(63b") (Vx) (x can do it D John can do it).
But the fact that we can differentiate the two any's by scope is not decisive,
since 64a and 65 are of the general form (Vx)(Fxza) where a is a constant,
and this is equivalent to the formula with an existential quantifier with limited
scope, (3x)FxDa. Thus we can rewrite 65 as
(65') ((3x) (John can sleep with x)) D (John will have a roof over his head)
24 There is a further ambiguity in the [+partitive] reading of 63b: 'If any one person can
do it, then John can do it (also)' vs. 'If any one person can do it, then John is that one
person.' The logical representations of these are quite different: (Vx) (x can do it D John
can do it) vs. (Vx) (x can do it D John is x). Both are examples of the extended [+partitive]
use of any, and both will be analysed as the result of the extended rather than limited scope
of any.
26 Although Reichenbach identifies the extended scope of any, as opposed to the limited
scope of every, he does not refer to that use of any which has the same limited scope as
every and is therefore responsible for the ambiguities noted here. He argues (106 ff.) for an
invariant interpretation of any with a categorical rule of maximally extended scope.
794
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
which we can paraphrase as 'If there is any girl that John can slee
he'll have a roof over his head.' We could therefore express the amb
as an operator with limited scope which is either a universal, [-par
tifier equivalent to every; or an existential [+partitive] quantifier, so m
the [+partitive] analysis. Simplicity considerations alone do not ea
the situation in favor of a conjunctive definition with variable sco
junctive definition with constant scope. But the variable scope pr
Savin also explains certain peculiarities in the restrictions on any w
and progressives to be examined below, and will allow us to simplif
statement of obligatory NEGATTRAC.
6.13. [-fact]. Any cannot be used in positive sentences with progr
preterits, but it can be used with modals, the general present, the
any tense which does not assert that an event actually occurred:
(67) a. *Anyone is going to his party.
b. *Anyone went to his party.
c. Anyone can go to his party.
d. Anyone goes to his party.
e. Anyone will go to his party.
f. Anyone might have gone to his party.
This set of properties can be summarized by saying that any is i
with predicates which make particular existential statements, ass
given factual state of affairs is the case. It is compatible with any l
general degree of assertion. I identified this feature as [-fact] in
continue to use this notation. In one sense, Anyone can go can be
statement of fact, since it can be factually disproved by finding s
cannot go. But it is neutral in regard to whether anyone in fact did
In ?2, it appeared that the obligatory character of NEGATTAC to
lieved by complementizers such as FOR-TO which are strongly selected
the [-fact] predicates identified by the Kiparskys and by Karttunen
anism of this connection remains to be elucidated; but if we are dea
same semantic feature, then any should be incompatible with [+fac
in general. The most obvious cases are
(68) a. *Anything is forgotten.
b. *Anything is finished.
c. Anything is possible.
d. Anything is available.
Here predicate adjectives containing perfective features, necessaril
incompatible with any, but those containing the semantic features
or modal suffixes are compatible. Adjectives neutral in regard to [
conflict with any:
(68) e. Anything is interesting.
The obverse of the [-fact] feature is that any is used to make la
ments about general conditions. Savin has pointed out that any m
28 The lawful generalizations expressed by any may be considered facts of a
not statements about a particular state of affairs. To make this point clear, S
feature [-mere fact].
795
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
tions about counter-factual conditions which every does not make:
(69) a. John spoke to everyone who came.
b. John spoke to anyone who came.
Here 69a is a factual assertion from which nothing else follows. But it follows
from 69b that if one more person had come, John would have spoken to him.
There is therefore a strong interaction between the quantifier and the predicate,
in that any is incompatible with a [+fact] predicate and also imposes a sense of
lawful generalization upon it.2 The feature [-fact] is therefore not a property
of any, but rather something required in the environment of any, i.e. a transfer
feature (Weinreich 1966:429). Transfer features can be seen operating in temporal
adverbs like ago which transfer a temporal feature to the noun phrase with
which it is in construction: symbolically [+temp -- (NP)]. The metaphor a grief
ago does not yield a re-interpretation of ago, but rather imposes a temporal feature
on grief. The transfer feature of any can thus be formalized provisionally as
[-fact -- (pred)], leaving open for the moment how we identify the particular
predicate phrase to be affected.
We noted, in 21-24, four sets of conditions which removed the obligatory
character of NEGATTRAC: a preceding negative, a preceding hypothetical, a following
relative clause, or embedding with non-finite predicates. All these also
operate to remove the constraint that any cannot be used with preterits and
progressives. The negative cases will be considered in the following section. In
the positive cases, we see that any in such environments is compatible with preterit
and progressive verbs:
(22') HYPOTaETICALS
a. If anyone went to his party, they had a good time.
b. When anyone went to his party, they had a good time.
c. Should anyone be going to his party, they will have a good time.
(23') RESTRICTED INDEFINITES
a. Anyone who was anybody went to his party.
b. Anyone who is anybody is going to his party.
(24') EMBEDDINGS
a. For anyone to have gone to his party was amazing.
b. For anyone to be going to his party is astonishing.
In all these examples, the [-fact -. (pred)] feature of any is transferred to some
other predicate higher than went or is going. They all have the general form
27 Reichenbach argues that any statements like Anything goes are best shown without the
universal quantifier as free variables, of the form Fx, and that this gives the support of
natural language to operations with such formulas by logicians. It would follow that any
is thus opposed to every and all as the absence of a quantifier, rather than as an indefinite
quantifier. But Reichenbach also states that Fx 'means the same' as (Vx) Fx, and this is a
necessary intermediate for his interpretation of any sentences with extended scope. Furthermore,
the distributive property of any is best shown logically as the ordering of quantifiers:
(Vx) (3y) Fxy instead of (3y) (Vx) Fxy. However, Reichenbach's argument for free
variables may be relevant to the ultimate analysis of our problem, since the negation of
Fx is not interpreted -((Vx) Fx) but rather (Vx) -Fx. This implies the possibility of a
further logical motivation for the constraint on All the boys didn't leave, Any boys didn't
leave etc., though the connection between the logical facts and the surface constraint is
certainly not clear.
796
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
(Vx) (Fx D Gx). The logical representation of both 22'a and 22'b wou
(70) (Vx) (x went to his party D x had a good time).
There is no difficulty in reducing 23' to the same form, since, as note
is generally agreed that this must be interpreted logically as equiva
there was anyone who was anybody, he went to John's party', giving
(71) (Vx) (x was important D x went to John's party).
Here the scope of the universal quantifier is the entire sentence. Si
antecedent and consequent are [+fact] predicates, we must conclude
not necessary for any to be compatible with either. The transfer fea
therefore apply to the if-then element of the sentence, which must be in
as the highest predicate if our [-fact -- (pred)] notation is to be ma
We would then have to establish the formal convention that the [-fa
is to be transferred to successively higher predicates (perhaps in the cour
cycle) until it agrees with one:
(72) I said that if John knew [-fact] that Harry invited anyon
party, he told me a lie.
The scope of any here extends up to the conditional clause:
(72') I said that (Vx) (John knew that Harry invited x D John to
le).
But if there is no predicate which will accept the [-fact] transfer fea
the use of any will be ungrammatical:
(73) *I said that John knew that Harry invited anyone to the par
he told me a lie about it.
If Savin's analysis of the ambiguity of any is correct, we should be able to find
only one meaning in 22'-24'. A limited scope for any, confined to the antecedent,
would be impossible because of the incompatibility of any with preterits and
progressives. This seems to be the case: we cannot maintain the 'every' meaning
of 64a if we change the antecedent to a preterit or progressive. Thus 64b' has
lost completely the meaning that John had slept with ALL the girls:
(64b') If John slept with any of these girls, then he had a decision problem.
This finding shows that the [?partitive] theory is inadequate to account for the
ambiguity of any in 63b and 64. If the scope of any were limited to the antecedent
in both 64a and 64b, there would be no way to explain that only one of these
meanings-the [+partitive] one-is compatible with a preterit or progressive
verb, as in 64b, and is the only meaning which emerges when we change the
antecedent of 63b to a preterit:
(74) If anyone did it, John can do it.
Since any is incompatible with did in 63b', its scope must necessarily extend
beyond the antecedent, and no ambiguity is possible. The scope explanation
proposed by Savin is the only one consistent with these facts.
The embedded sentences of 24' show an extended scope for any. It is possible
to interpret these sentences as if-then conditions, and it seems likely that it is
necessary for any to be compatible with the main predicate-in the case of
amazing, unlikely etc. These involve negative presuppositions, that 'someone did
not believe that anyone would go'. The problem of apparently neutral predicates
797
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
like O.K. with me, raised in 26 in connection with NEGATTRAC, seems to involve
a more indirect type of presupposition: 'for someone it is not O.K. for anyone
to go', but for me it is. An exact formulation of the conditions which permit any
in these and parallel constructions is beyond the scope of this paper, though not
presumably beyond the scope of any paper. They seem to involve all the semantic
features we have identified in any and perhaps more.
We can sum up our characterization of any in positive sentences by paraphrasing
any X as a prediction of the form:
(75) If [-fact] you select [+quantifier], one at a time [+distributive], all
[-partitive] of the items designated X, no matter which [-specific]
or how many [-definite], you will find that...
The combination of [+distributive, -partitive] is usually realized as each, so
that we can reduce this to 'If you select each of the items designated X, no matter
which...' Any is thus opposed to each in its indefinite feature, to all by its
distributive and indefinite features; and since every is [? distributive], its relation
to any may be the same as all or each. The transfer feature [-fact] is not needed
to distinguish these quantifiers minimally:
(76) UNIVERSAL [-partitive] QUANTIFIERS
[ + definite] [- definite]
[+distributive] each everyi any
[- distributive] all every2
In ?2, we observed that stress on any, with or wit
diminished the obligatory force of NEGATTRAC. It is th
that stress on just any has no such effect in eliminat
using any with preterits and progressives:
(77) a. *Anyone is going to his party.
b. *Just anyone went to his party.28
6.2. Any WITH NEGATIVES. We now turn to the be
sentences, and reconsider the obligatory rule of NEGATT
stand its strongly obligatory character in terms of the
isolated. There is a general output constraint (28
heterogeneous list of conditions under which NEGATTRA
we find that (I) NEGATTAC is uniform in all dialects
do not occur in natural speech, and (III) violations co
are not interpretable and do not become more accepta
our business to explain here the existence of the ru
result of the interaction between the properties of th
have developed over the course of time. But in prop
contrasts sharply with negative postposing and NEGC
of this paper is on the contrast between categorical
variable rules, and our analysis is designed to illumi
We noted above that the first generative formulatio
formal, suppletive variant of some in negative senten
28 Although Just anyone went is not acceptable in isolation,
amplification of a negative response to Did anyone go to his
Here the mechanism seems to involve the overt denial of t
that it is not the case that it does NOT matter which of the members of the class are selected.
798
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
the following series:
(78) a. I want some.
b. I don't want any.
c. *I want any.
d. *I don't want some.
A careful reading of Klima's original discussion shows that he does not explicitly
mark sentences of type 78d as ungrammatical, and many writers have since
pointed out that they are quite acceptable in a wide range of contexts. As noted
above, R. Lakoff has shown many meaningful alternations of some and any in
questions.
There is no shortage of empirical evidence to show that 78d is grammatical.
Among our BE-speaking subjects, we find many examples of some in negative
sentences: I don't want some more (Keith W., 9, Thunderbirds); I don't give my
friends some (David H., 13, Thunderbirds); We don't sometime use that (Kenneth
S., 12, New York City). What then is the basis of the feeling behind Klima's
formulation which marks I don't want some (78d) as at least awkward or unlikely
in unmarked contexts, and I don't want any (78b) as much more natural? It
seems to me that those who declare 78d ungrammatical are unconsciously treating
it as the CONTRADICTION of 78a, and the full contradiction of an indefinite
existential statement is a universal negative.
Consider someone looking out of a window who suddenly says, There are some
rabbits in that field. If a friend comes to the window, he cannot deny the statement
as quickly as it was made. Since the location of the rabbits is indefinite, he
must now scan the entire field before he observes, There aren't any rabbits in that
field. If he takes a quick look at one corner of the field and hastily observes,
I don't see some rabbits in the field, then he has made a true observation, but he
has not contradicted the first statement. It is necessary to search the entire
field, and obtain class closure to contradict an indefinite statement.29 This is
shown diagrammatically as follows:
some NEG + any
The correspondence of so
siderations, rather than a
29 The strong denial with [-p
some in the same way that a s
denial with the preterit. Yo
meaning'at least once', but d
times'.
799
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
course strategy which established the sequence of 78a-b, rather than an obligatory
rule of grammar. If the any in 78b were simply a formal alternant of some,
we would expect to find the semantic features of some, such as [+partitive]. But
by the argument advanced above, any is selected just BECAUSE it contrasts with
some in being [-partitive]. We have every reason to believe that any in negative
sentences is the same universal quantifier that we have just examined in positive
sentences.
It is now generally understood that the negation of sentences with quantifiers
can take several forms, and that some surface structures with quantifiers and
negatives are ambiguous in this respect. Carden explains the two interpretations
of All the boys didn't leave by readings of the deep structure with either the
quantifier as the highest predicate (NEG-V) or the negative as the highest (NEG-Q).
Logically, these two interpretations would be opposed as
(79) a. NEG-V (Vx) ({Fx)
b. NEG-Q ( fVx) Fx.
Carden finds that some subjects get both readings, some only one; but for everyone
NEG-V is rendered more naturally by None of the boys left. There seems to be
general agreement on the following series for (Vx) (rFx) readings:
(80) a. All the boys didn't leave.
b. Every boy didn't leave.
c. ?Each boy didn't leave.
d. *Any boy didn't leave.
While 80a-b are awkward but grammatical for most speakers, 80c is questionable
or more than questionable, and, as we have seen, 80d is impossible for everyone.
If the [+distributive] feature of each is responsible for 80c being worse than 80a,
then 80b should be intermediate; some subjects should react to the [+distributive]
alternant of every. And in fact preliminary investigations of this series suggest
that this is the case: 80a-d form an increasingly ungrammatical series. For
the NEG-V meaning, all these forms are replaced by None of the boys left, or No
boys left; I have advanced reasons above for believing that NEG + any underlies
no and none here. NEGATTRAC therefore seems to apply with moderate strength
to all and every, strongly to each, and categorically to any. From our analysis of
the feature system of these quantifiers in 76, it would seem that it is the interaction
or cumulative effect of the non-partitive, indefinite, and distributive character
of any which is responsible for the force of NEGATTRAC here. That is,
[-partitive, -distributive, +definite] + NEG is awkward (all, every); [-partitive,
+distributive, +definite] + NEG is objectionable (every, each); and [-partitive,
+distributive, -definite] + NEG is impossible. The total effect of Any
boys didn't leave is that of going up an indefinite number of garden paths, one
after the other, and knowing all the time that there is nothing at the end of any
of them. It should be borne in mind that a logical equivalent of the universal
negation (79a) is the existential negation
(81) (-3x) Fx
which is parallel to the equivalent surface structures:
(82) a. None of the boys left.
800
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
b. No one of the boys left.
c. Not one of the boys left.
d. There was not one of the boys that left.
e. There wasn't any boy that left.
In all these versions of 81, the negative precedes the indefinite and
satisfies constraint 28. It is not likely that there is a one-to-one corre
of logical representations 79 and 81 with surface structures 80 and 82.
less, the parallel is so striking that we might rewrite 28 as follows:
(28') If a linguistic structure is interpreted as a positive universa
fier applied to a simplex negative predicate, then the neg
will precede or be incorporated with the quantifier in the
structure-often when the quantifier contains a distributiv
ture, and always when it is indefinite.
This new constraint is broader than 28 since it includes all the quantif
But it seems to lack the list of exceptions to the categorical operation
ATRAC on any-that NEGATTRAC may not apply when the indefinite
manded by another negative, (b) commanded by a hypothetical, or (c)
We can now use our analysis of any in positive sentences to show th
specifications are not needed in 28':
(a) Commanded by another negative: when a second negative comm
indefinite quantifier, as in 19, I don't say that anyone didn't go, the logical
is not that of 28', but rather
(83) r-((Vx) -FFx)
-that is, 'I say that it is not the case that anyone didn't go.' In 19, n
raising is added to 83: the effect of a preceding negative without negative
is seen in ex. 3 of Table 1. Such explicit negatives as it is false do relieve th
gorical nature of NEGATTRAC; but, as noted in ?2.1, they are not as ef
the implicit negatives in regret, apologize, am sorry, hate etc. It is not cle
these indirect negatives have a stronger effect.
(b) Commanded by a hypothetical: these are the cases listed under
and analysed above (?6.1) in their positive counterparts. From our di
it is clear that the logical structure of 22, If anyone can't go ... etc.,
Anyone can't go who drinks etc., are not of the form (Vx) -Fx, but ra
(84) a. (Vx) (-Fx D Gx)
b. (Vx) (-Fx = y).
Neither of these is specified in 28'. In the case of the embedded sentences of 24,
For anyone not to go ... etc., the structure is either 83 or 84b, and it is difficult
to decide which. That is, 24b, It's a shame for anyone not to go, may be read as
(85) (Vx) (3y) (-(x go) D shame y)
or we can refer to the presupposition of 24b: 'someone doesn't want anyone not
to go':
(86) (Vx) (3y) (-(y like -(x go))).
(c) Stress: in all the previous cases cited, the scope of any naturally expands
to cover the entire proposition. We observe that, in 17'c, Just dnybody can't eat
801
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
there (?2.2), the stress protects against the force of NEGATTRAC. But it is also
clear that the meaning is quite different from Nobody can eat there. The logical
representation would be equivalent to the NEG-Q reading of 79b:
(87) (-YV) (x can eat there)
where the negative has a larger scope than the quantifier. This is not the normal
scope for any; the limited scope shown here is more characteristic of all and
every, as in Not every boy can eat there. But the stress on any has the effect of
limiting its scope, so that the negation applies directly to the quantifier.30 There
is no need for any special representation of the effect of stress on any; contrastive
stress normally focuses the semantic force of negation on the stressed particle so
that the negative commands it. In
(88) The ]i string didn't break
we assert that it was NOT the E-string but some other string which broke. And
in 17'c we assert that it is not ANYbody who can eat there, but only members of
some other, less inclusive, set.
The mechanism of contrastive stress involved here does not operate without a
negative feature. It is therefore understandable that the parallels between the
constraints on any in positive and negative sentences come to an end here, and
that stress on any does not make it compatible with preterits and progressive
verbs.
6.3. TEE LOGICAL REPRESENTATIONS. Though we should not attempt to make
any direct connection between the logical representations and the relatively
superficial rule of NEGATTRAC, it is evident that this transformation is governed
by both logical configurations and abstract conceptual features of the sentence.
These considerations are quite distinct from the emphatic, affective, and socially
marked features of negative postposing and NEGCONCORD, and fully justify our
treating NEGATTRAC as a separate transformation.
Although the informal logical representations given above are quite revealing,
it is clear that they are not sufficient to show the meaning of any. Any is more
than a universal quantifier; we would have to enlarge our representation of
Anyone can go to read as follows:31
(89) (Vx) (x can go)
[-fact -- (pred)]
[+distributive]
30 In the representations used by Carden, this difference would be shown as a deep structure
with the quantifier as the predicate of the highest level sentence, as opposed to one
with the NEG as the predicate commanding the quantifier.
' That is, features such as [+distributive], [-definite], [-fact -> (pred>] are not shown
in the logical representation. As noted above, [+distributive] can be shown logically as the
difference between (Vx) (3y) Fxy and (3y) (Vx) Fxy, but not for one-place predicates such
as Anyone can go. The [-definite] feature may be shown by the absence of a quantifier, as
indicated in fn. 27. The non-factual transfer feature might ultimately be reduced to a consequence
of this free variable formula. Whether or not all the semantic features of the quantifiers
can be reduced to a logical representation is still an open question; an answer will of
course require a proof of such suggestions with an explicit connection between logical formulas
and the grammatical end result.
802
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
We thus have three levels of semantic features operating in such se
There are features such as [+quantifier, -partitive] which can be co
accounted for by logical notations; there are abstract, conceptual fea
as [-fact] which must be added to the two-valued logical notation; a
are affective features such as emphasis which have no place in the logical
Given the rules of English, we expect that an unstressed, unmodifi
indefinite introduces a positive sentence. Violations of NEGATTRAC, e
can't go, thus begin with what should be a positive sentence and the
the hearer with a negative. Without the presence of a hypothetical f
modifying clause which may be interpreted as a hypothetical featur
stress to limit the scope of any, the sentence remains an uninterpretable
of our expected pattern. It is possible that all universal and compellin
tions of this type involve abstract conceptual features with unexpecte
of meaning. In any case, we have no choice but to regard NEGATTRA
tirely different rule from the rightward movement of the negative
clarify further the distinction between obligatory and variable rules, we
consider the way in which negative postposing and NEGCONCORD op
see what semantic or discourse functions can be assigned to these transfo
7. THE NATURE OF NEGCONCORD. The semantics of NEGCONCORD is a much
simpler study than that of NEGATTRAC. To begin with, we note that NEGCONCORD
is an optional rule for almost all dialects of English. In this respect, it resembles
negative postposing of the standard language. Both represent rightward
movement of the negative beyond the verb, and they seem to share a strongly
emphatic character:
(90) a. I owe you nothing!
b. I don't owe you nothin'l
Jespersen points out (1924:331) that the system of 90b is the normal one for
most languages: negation is typically cumulative, not multiplicative. It is the
cumulative character of NEGCONCORD, along with the postposing, which seems
to put strong emphasis on the negation itself. It is not necessary to base such
32 The property of any which allows the interpretation of modifiers as hypotheticals is of
considerable interest, and may be important for the exploration of the semantic value of
various nominalizing processes of sentence reduction. We can construct a series of sentences
with decreasing acceptability:
Any boy who received an invitation went to his party.
?Any boy receiving an invitation went to his party.
?Any boy with an invitation went to his party.
?Any boy invited went to his party.
?Any boy on his block went to his party.
?Any boy here went to his party.
*Any boy went to his party.
A purely logical representation will show the last item as a general hypothetical, as readily
as the first: (Vx) (x is a boy D x went to his party). But the grammar seems to reject this
interpretation, just as it would reject the interpretation of *Anyone went as (Vx) (x is
[+human] D x went). If we accept this extremely analytical approach, we must conclude
that any indefinite expression which contains a semantic feature in addition to the quantifier
is to be interpreted as a hypothetical, and logical interpretation would be at the mercy of
the number of distinctions made in the pronoun paradigm.
803
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
observations on intuitive speculation. When the pattern typical of careful speech
gives way to the vernacular, NEGCONCORD is used by speakers to make their
strongest points stronger. The use of NEGCONCORD can be illustrated by an
interview with a 60-year-old woman from a working-class family in East Atlanta,
Georgia. Mrs. Gratton had only a grade-school education, but she was self-educated
to the point of controlling most SE features when she wanted to. In the
entire face-to-face interview, she used the standard rules for negatives with indeterminates
32 times (including three examples of negative postposing); she
used NEGCONCORD only twelve times. For the first twenty minutes, Mrs. Gratton
regularly used such standard forms as the following:88
Interviewer: How do you keep it [the skillet] from getting rusty?
Mrs. G.: Well-uh-I haven't ever thought anything about it. I jus' wash it an' put it
back [laugh], an' clean it and that's all.
IVer: Do you keep it on top of the stove?
Mrs. G.: No-I usually put it up. Now you see, I haven' put up anything.
But a little later, the conversation touched on a theme that produced a sudden
switch to NEGCONCORD:
IVer: Do you make 'em [biscuits] from scratch?
Mrs. G.: Make 'em from scratch [chuckle].
IVer: Wow! Do you measure the things when you putMrs.
G.: I don't
Mrs. G.: measure nothin'! I never have
IVer: Never, even when you first ...?
Mrs. G.: measure' nothin'. I have never measuredThe
emphatic character of Mrs. Gratton's response shows up in the repe
strident overlap of speakers as well as her repetitions on the same theme.
this theme-cooking without measuring-comes up later in the interview,
again given this emphatic treatment. (Note that the cultural feature [-bo
learning] coincides with the use of non-standard grammar at this point,
reinforces the use of NEGCONCORD.) As the discussion continues, Mrs. Gr
switches back to standard forms:
IVer: How'd you learn how to make biscuits?
Mrs. G.: Well ...
IVer: Do you remember?
Mrs. G.: No I didn'. I didn' know how to do anything when I cooked. But-I could make
'im corn bread. That's about th'easies' thing anybody can do, can make corn
bread; i's nothin' to that.
NEGCONCORD in other white dialects typically has this emphatic character:
(91) We wouldn't never have took it! (Marie C., 39, Hackney, London)
(92) You don't have to know too much of nothing. (Monnie O., 35, Wanchese,
N.C.)
(93) He [my father] didn' take nothin' offa nobody. (Henry G., 60, E.
Atlanta, Ga.)
In all such dialects, NEGCONCORD is optional. Mrs. Gratton's husband used
3 I am indebted to Teresa Labov, who was the interviewer here, for initiating this ex-
change.
804
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
NEGCONCORD more than she did-17 times-but did not use it 12 times in the
face-to-face interview. In group sessions within the family, the percentage of
NEGCONCORD rises, but never predominates to the exclusion of the any forms.
This confirms our view that any with the negative does not contain the feature
[+neg] as some linguists have suggested, since it always contrasts with NEG +
any.
7.1. NEGCONCORD IN BE. The first thing that we note in BE is the extraordinary
proliferation of the negative:
(94) Once you get an even break, don't fuck it up, cause you might not
never get no time see 'im again. (Larry H., 15, Jets)
(95) I ain't never had no trouble with none of 'em. (William T., 25, Fla.)
(96) You better NOT never steal nothin' from me. (Jesse H., 16, Jets)
(97) He ain't not no Collins. (Speedy J., 15, Cobras)
The expansion of the negative goes beyond the two environments INDET and
VB, since in 97 two negatives are associated with the verb. Not can occur in
foregrounded position without the attraction of an indeterminate:
(98) I can walk through this wall, but not my physical structure can't
walk through this wall. (Stanley K., 15, Cobras)
This development can be related to elliptical not, which negates noun phrases,
adverbs etc. without verb phrases, as in standard not exactly:
(99) Not no more. They don't fly no more. (Lawrence W., 15, Jets)
(100) Not none-Not none o' our friends, but BIG people. (David H., 13,
Thunderbirds)
This negative is realized from the sentence modality without going through the
morphophonemics of pre-verbal location. The fact that it can do so means that
we cannot allow NEGCONCORD to be dependent upon a pre-verbal transformation.
There is also a metalinguistic not, used to edit texts and correct mistakes.
This too can generate NEGCONCORD:
(101) I went to some f-not no friends of mine, but a friend of MY friend.
(Pauline J., 37, S.Car.)
Another aspect of the exuberant use of the negative in BE is NEGCONCORD
referring to a negative not contained in the sentence at all, but contained in a
presupposition or implication of the sentence. This 'NEG from nowhere' shows
up in a toast, 'The letter'. A whore who left the hero to rot in jail asks him for a
dime:
(102) Bitch, before you get the price o' NOTHIN',
A grape got to grow as large as a pumpkin.
All dialects normally have anything in clauses introduced by before, unless, etc.
The implication of 102 and all that follows is that she won't ever get any
money. We also observe this abstract NEGCONCORD in
(103) She always gets kids in trouble for nothin' they didn' do. (Vineland,
N.J.)
Here the standard equivalent is not immediately obvious: 'She always gets kids
805
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
in trouble for things they didn't do.' The fact that those things are NOTHING
from the standpoint of the kids seems to influence the NEGCONCORD.34
The most relevant fact about NEGCONCORD in BE is that it is NOT optional;
in the major environment, within the same clause, NEGCONCORD to indeterminates
is obligatory. Before we examine the evidence for this statement, the
relevant environments must be strictly defined and a number of distinctions
made:
(I) NEGCONCORD is never obligatory to the pre-verbal position. We can easily
locate BE sentences where the negative is NOT transferred to pre-verbal position:
(104) Nobody was after 'im. (Lawrence W., 15, Jets)
(105) Nobody fights fair. (Henry N., 29, Bronx)
(II) The indefinite article a is not an indeterminate and is not involved in
NEGCONCORD:
(106) a. I ain't never lost a fight. I ain't never lost a fight. (Robert, 13,
South Carolina)
b. Ain't nobody a man. (Stanley K., 15, Cobras)
c. You ain't got a funny bone. (Michael, 13, W. Philadelphia)
From evidence given in various sections of this paper, we conclude that the
underlying form of no is NEG + any, not NEG + a, which is realized as not a:
(107) I'm not a baby. (Pam, 13, W. Philadelphia)
(III) Forms appended as sentence modifiers, like either or anything, are not
to be considered within the same clause; they are to be classed with NEGCONCORD
outside the sentence with NEG. Some of these show concord:
(108) I didn' know nothing' about the people, or nothin'. (Pauline J., 37,
South Carolina)
(109) I might can't get no more fines, neither. (Thomas H., 11, New York
City)
But very often we find no NEGCONCORD to such positions among vernacular
speakers of BE:
(110) They're not too hip, or anything. (James T., 13, Jets)
(IV) The obligatory character of NEGCONCORD within the clause is not at
all similar to the obligatory nature of NEGATTRAC. Any speaker is potentially
capable of omitting the rule, and so producing sentences with any. He hears
the standard form, and can interpret it, and in his careful speech he usually
shifts away from the 100 percent use of NEGCONCORD. Even in casual speech
many adults have shifted away from the BE vernacular and lost consistency in
NEGCONCORD. And most importantly, consistent use of NEGCONCORD is the
characteristic of core speakers of BE in their peer-group interaction. Marginal
members of the peer-group culture and isolated individuals ('lames') do not
show consistent NEGCONCORD.
Table 2 illustrates the four restrictions just outlined on the obligatory character
of NEGCONCORD with evidence gathered from a large body of natural speech.
This table shows five peer groups of BE speakers. We show figures from the
34 This example is not from BE, but from a Puerto Rican family in Vineland, N.J., who
had minimal contact with the BE of New York City and Philadelphia.
806
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
PROPORTIONS
OUTSIDE THE
PROPORTIONS WITHIN THE CLAUSE CLAUSE
TO
INDETERTO
INDETERnINATES TO VERBS MINATES
STYLE INSTANCES SPEAKERS INSTANCES INSTANCES
I II III IV
T-Birds Group 62/62 5/5 2/13
(9-13 yrs.) Ind'l 94/96 11/13 12/34 1/11
Cobras Group 70/70 11/11 12/35 0/1
(12-17 yrs.) Ind'l 186/188 14/16 7/18 0/5
Jets Group 149/151 13/15 10/38 1/3
(12-16 yrs.) Ind'l 360/370 25/30 39/97 2/7
Oscar Bros. Group 53/55 4/5 10/21 5/5
(15-18 yrs.) Ind'l 79/81 3/4 6/12 0/2
Danger Girls Group 48/48 4/13
(13-17 yrs.)
Lames Ind'l 73/81 10/12 1/6 0/3
Inwood (white) Group 25/32 2/7 0/36 0/2
(11-16 yrs.) Ind'l 34/42 2/8 0/35 0/1
TABLE 2. Use of the NEGCONCORD rule by speak
Columns I, III, and IV indicate the proportion of
number of possible instances.
Column II indicates the proportion of informant
out of the total number of informants.
first four groups for NEGCONCORD in two styles: group and individual. The individual
interviews were often lively, carried on with interaction from other
members of the group; the interviewer (John Lewis) was a member of the vernacular
culture himself. Nevertheless, the individual interviews cannot be taken
as representative of the most casual style, where the minimum attention is given
to speech (Labov 1970). As long as questions are being asked and answered, we
consider the style used not to be the basic vernacular as just defined. The group
sessions provide the context which defines the vernacular-where speech is
controlled by the same factors operating in everyday life (Labov et al., I:57-64).
The first row of Table 2 shows that in group sessions, five members of the
Thunderbirds spoke 62 sentences in which NEGCONCORD might have been used,
and all speakers always did so. But in only two of thirteen cases was the negative
transferred to the verb in the same clause. In individual sessions, concord was
used in 94 of 96 possible cases. Eleven of the thirteen members scored 100 percent,
but two members each used the standard form once. Again, transfer to the
verb took place at a much lower frequency-12 of 34 cases-and in only one
of 11 cases was the negative transferred to an indeterminate in a following clause.
The older groups show essentially the same pattern. The Cobras again show
100 % NEGCONCORD, in group sessions, to indeterminates within the clause. In
the case of the Jets, there are two cases in group sessions where one individual
uses a standard form once. The Oscar Brothers, who are beginning to move out
of the BE vernacular culture, show a slightly greater tendency in this direction.
807
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
In all cases, the frequency of transfer to the verb is much lower than to indeterminates,
and transfer outside the sentence is irregular.
The 'lames' are isolated individuals interviewed in vacation day centers and
on the streets of the Jet and Cobra areas, where it was known that they were
not members of the central peer groups. Their use of NEGCONCORD is slightly
less than that of the peer groups, though the difference is by no means as striking
as in other indices of the BE vernacular (Labov et al., II:4.4). The white
Inwood figures are markedly different: even with the small numbers of instances,
we find that only two speakers use NEGCONCORD all the time, and of course there
is no transfer of the negative to the verb. Other studies of NEGCONCORD among
white speakers in New York City (Labov 1966) confirm this pattern.
Turning to the evidence of repetition tests with speakers of BE, we find a
pattern of response which reverses the pattern seen with violations of NEGATTRAC.
Sentences with or without NEGCONCORD are understood immediately,
and repeated usually with NEGCONCORD. Here are some responses to a sentence
without NEGCONCORD:
(111) Nobody ever took an airplane, and none of us took a bus, either.
Boot: N-N-Nobody never took a airplane, none of us took a bus,
neither.
Money: None of us never took an airplane, and none of-take-none
of us never take a bus, either.
Among the adolescent Jets, eight of sixteen sentences with negative plus ever
were repeated with never, though a much smaller percentage of sentences with
any was corrected to none. On the other hand, the majority of all BE NEGCONCORD
forms were repeated without change. Sentences modified with anyhow
or nohow were repeated without NEGCONCORD about 75 % of the time (Labov
et al., 3.9).
Information from Shuy et al. 1967 fits in with this pattern. That study was
limited to individual interviews, like the original New York City study of
Labov 1966. The interviewers did not have as much success in identifying and
separating casual speech from careful speech within these interviews, so that
all the conversational parts are taken as a single style. Wolfram 1969 analyses
the data for 48 Black speakers in the Detroit study, and includes NEGCONCORD
in his data. In each social class, he selected four pre-adolescents, 10 to 12, and
four adolescents, 14 to 17 years old. In the Lower Working Class, four of eight
showed 100 % NEGCONCORD, and in the Upper Working Class, three of eight.
If we compare these figures for the individual interviews in Detroit with our
individual interviews by a Black participant observer, the Detroit study shows
a similar semi-categorical use of NEGCONCORD in the BE vernacular. The Detroit
study also agrees with our findings in the restrictions on NEGCONCORD.
None of the adults show 100 % use of this rule. There is much less use of the
rule with indefinites outside the clause, or in pre-verbal position; and if the
indefinite article a is included in the analysis, the percent of NEGCONCORD
drops sharply. Other data support the notion that the indefinite article is not
to be included as a site for NEGCONCORD.
For a further body of independently gathered data, we can turn to a series of
808
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
PROPORTIONS
OUTSIDE THE
PROPORTIONS WITHIN THE CLAUSE CLAUSE
TO INDETERTO
INDETERMINATES TO VERBS MINATES
INSTANCES SPEAKERS INSTANCES INSTANCES
I II III IV
Girls (9-13 yrs.) 75/76 9/10 0/1 5/7
Boys (9-14 yrs.) 70/74 7/10 1/4 3/9
TABLE 3. Use of NEGCONCORD by two group
in West Philadelphia.
recent observations made of pre-adolescent
by Marjorie Goodwin.36 The groups were re
observation, and the language recorded is t
activities-making slingshots, jumping rope
vernacular culture actively controlling each
content, these samples of Philadelphia spee
recorded in the group sessions in South Ha
shown in Table 3. Both groups come close t
the girls, there is only one use of a standar
that fortune-telling can be the cause of a d
leader of the group, says:
(112) No. I don't want anybody to die in
The use of a standard form in a solemn context is a common form of codeswitching,
and supports the view that NEG + any is in another system as far
as Pam is concerned.86 A little later, she says:
(113) They ain't nobody died in my family.
Among the boys, there are a few more instances of the use of standard forms,
mostly by older members of the group. The majority, however, use consistent
NEGCONCORD to indeterminates within the sentence.
The use of NEGCONCORD outside the original sentence is less regular, but we
do not have adequate data to distinguish various cases. In the Philadelphia
group, the negative is consistently transferred to indeterminates in lower sentences
commanded by the original negative:
(114) Y'all ain't gonna be able to make no slings till we get all these slingshots.
(Poochie, 13)
(115) She didn't mean that to do nothin' to you. (Maria, 13)
(116) I ain't know I could do none of that. (Poochie, 13)
At the same time, the negative is less often transferred to reduced clauses or
8a I am much indebted to Mrs. Goodwin for the use of these materials, which she gathered
and transcribed in the course of research for the Center for Urban Ethnography at the
University of Pennsylvania. Her analyses of personal interaction have also provided the
insight necessary for a correct semantic interpretation of the sentences cited. In some cases
only isolated sentences are quoted, but the interpretation is derived from the extended context
of group interaction.
36 There is a close parallel here with the form of code switching used in Hawaiian Creole
in solemn narrations. The -ed past does not appear in the Creole, but is imported for solemn
effects in such utterances as He died (Labov 1971b).
809
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
sentence modifiers:
(117) You can roll your eyes all you want to 'cause I'm tellin' you-tellin',
I'm not-not askin' you. And I ain't say no please, either. (Pam,
13)
(118) They ain't hardly even gonna be able to load 'em that fast either.
(Tokay, 13)
This extra-sentence either can contrast with the indeterminate either within
the original sentence:
(119) We don't want neither one of y'all. (Michael, 13)
We thus find that the use of the NEGCONCORD rule for the West Philadelphia
group is identical with that used in Harlem, and coincides with additional data
from exploratory interviews in Detroit, Cleveland, Chicago, and Los Angeles.
Since NEGCONCORD is practically obligatory in the most important case, we
must raise the question: in what way can it be said to be an emphatic trans-
formation?
7.2. THE EXPANSION OF EMPHATIC NEGATION IN BE. We have observed that
the standard use of any + NEG is available for most BE speakers, but primarily
as an importation from an outside system. Since NEGCONCORD is practically
automatic within the original sentence, it cannot add any emphasis, as it did
for Mrs. Gratton. It is simply an automatic adjustment of the grammar. For
Mrs. Gratton, I don't measure nothin' contrasts with I don't measure anythin'
by the addition of an optional [+neg] to the final lexical item. If such an addition
is obligatory, it cannot contrast with its absence. How then do BE speakers
convey the equivalent of the emphatic contrast registered by Mrs. Gratton?
There are a number of answers to this question, which we can illustrate with
quotations from the West Philadelphia group. First, the BE vernacular often
adds negatives beyond the provisions of the NEGCONCORD rule. Even within the
rule, it provides for negatives in pre-verbal as well as in subject position. But
indeterminate subjects are relatively rare. In most cases, NEGCONCORD operates
from the verb rightward to an object indeterminate or to an indeterminate
noun phrase modifier. Both cases appear in
(120) I ain't write nothin' in no street. (Pam, 13)
Sometimes an indeterminate modifier is added when there is really only one
semantic object to focus on:
(121) She ain't in no seventh grade. She in eleventh grade. (Maria, 13)
The NEG + any which modifies seventh grade can only be understood as adding
emphasis, since there is only one seventh grade to be considered here. That is,
Maria is not reviewing all the possible seventh grades that the girl could be in,
but is rejecting the notion that she is in the seventh grade at all. Thus She
ain't in no seventh grade contrasts with She ain't in the seventh grade as a more
emphatic denial. Similarly, we can observe a contrast of the indefinite article a
with the NEG + any quantifier in
(122) I don't want a piece. I want a whole one. I don't want no piece.
(Poochie, 13)
810
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
Here Poochie strengthens his first negation with the insertion of the u
quantifier any which automatically attracts the negative. Emphas
plished by substituting the universal [-partitive] quantifier any fo
tial [+partitive] pronoun a. The NEGCONCORD rule heightens the e
it automatically adds [+negative] to reinforce the contrast.
Such postposed negatives can be heard as emphatic even when NE
is obligatory, if they are moved further rightward. Florence of West P
follows up Pam's denial in 120 with a more emphatic form:
(123) We ain't write over no streets nothing.
Here the extraposition of nothing to the end of the sentence has the s
effect as standard negative postposing, We wrote nothing over any str
Even though negative concord is automatic in BE, we see that t
number of resources for reinforcing negation in the original sent
troduction of extra quantifiers, postposing negatives, transfer to
position, and reduplication of free not. In addition to these devic
also extended the standard rule of negative inversion to give additi
to sentences with indeterminate subjects. The standard rule as us
black and white speakers preposes neither, not, or adverbs containi
feature:
(124) The Negro doesn't know about the Negro, and neither does the white
know about the Negro. (Byrne C., 26, Ohio)
(125) Not until he came into United States did-uh-they decide to get
married... (Charles I., 31, Bronx)
In many Southern dialects, and in BE, the tense marker and negative can be
inverted with an indeterminate subject, yielding
(126) Doesn't nobody really know that it's a God, you know. (Larry
H., 15, Jets)
(127) Don't no average motherfucker make no fifty dollars a day. (Vaughn
R., 16, Jets)
(128) Don't nobody break up a fight. (Willie J., 15, Chicago)
Along with the tense marker, the first member of the auxiliary can also be moved,
just as in the flip-flop rule for questions. Thus negative inversion can involve
can't, wasn't, won't, and ain't. A study of negative inversion would require a
separate paper, but the most important feature to note here is its strongly emphatic
character:
(129) Didn't nobody see it, didn't nobody hear it! (Benjamin S., 45, N.Y.C.)
(130) Ain't no white cop gonna put his hands on me! (Jesse H., 16, Jets)
(131) Ain't nobody in my family Negro! (Boot, 12, Thunderbirds)
(132) I know a way that can't nobody start a fight. (Willie J., 15, Chicago)
(133) Won't nobody catch us! (Willie J.)
Negative inversion plainly depends upon and follows NEGTRANSFER from subject
indefinite to the verb, i.e. Nobody will catch us -- Nobody won't catch us --
Won't nobody catch us. Six cases of negative inversion in West Philadelphia
should accordingly be added to the figures in Table 3 on NEGTRANSFER to the
verb by rule 53, giving 3/4 for girls and 4/7 for boys. One could argue that the
811
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
movement of the negative to the left of the indefinite is a continuation of the
NEGATTRAC rule, and is reinforced by the factors outlined in ?6. But negative
inversion is always optional, and it can be seen in these examples to have the
same emphatic force as the NEGCONCORD rule.
When negative inversion moves an auxiliary like won't as in 133, it must be
derived from an underlying sentence with only one tense marker (nobody will
catch us). But when ain't precedes the subject, there is often the possibility that
it originated in a clause of its own with deleted dummy it:
(134) Ain't nobody know about no club. (William T., 25, Florida)
This could be based on either of the following:
(134') (It) ain't nobody (that) know about no club.
(134") Nobody ain't know about no club.
This is a possible route for 130 and 131 as well. Note that these cases require
deletion of relative pronoun that in subject position; such deletion seems to be
more common in BE and in Southern dialects than elsewhere.37 On the other
hand, 135 can only be based on 135' and not on 135":
(135) Ain't nothin' you can do about it.
(135') (It) ain't nothin' (that) you can do about it.
(135") * Nothin' ain't that you can do about it.
We occasionally have sentences that are syntactically ambiguous, like 134,
and also contain a negative in the second clause. One interpretation of such
sentences is that they are not derived by negative inversion, but by it deletion,
and therefore show NEGCONCORD to a verb in a following clause:
(136) Ain't no cop never beat me in my head. (Jesse H., 16, Jets)
7.3. EXPANSION OF NEGCONCORD TO OTHER QUANTIFrmfiS. There is one final
way in which the use of emphatic negation has expanded in BE. Transformations
which were originally limited to indeterminates any, ever, and either can occasionally
apply in the presence of other quantifiers like many and much. In 1,
47, and 48, we see that the rule which extends NEGCONCORD to the verb in following
clauses requires an indeterminate in the first clause; but in 50, this rule
operates with a subject much. This extension should not be entirely unexpected;
as Baker 1970 has pointed out, much has negative polarity: He didn't say much
but *He said much. The same cannot be said of many, however. We find negative
inversion operating with subjects containing this adverb:
(137) It's against the rule; that's why don't so many people do it. (Willie J.,
15, Chicago)
(138) Don't many of them live around here. (Hough area, Cleveland)
And in 49, the citation from 'The fall', the extension of NEGCONCORD to a following
clause is triggered by an indefinite with a, which is not a part of the
usual NEGCONCORD rule and does not incorporate the negative in the way that
87 E.g., There is a lot of women wanna make you do more than just hit 'em (Weldon Sorrell,
46, N.Car.); An' tha's the rules mean (Boot, 12, T-Birds); I got a man to come out here Saturday.
Is gon' paint (Henry G., 60, E. Atlanta), We have very few go to college (Monnie 0., 35,
North Carolina).
812
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
any does: For there wasn't a son of a gun who this whore couldn't shun. W
fore see speakers of BE extending the scope of the NEGTRANSFER ru
vironments beyond any formal rule that we can now write, since t
reason to believe that these are productive patterns.
8. AN OVER-ALL VIEW OF NEGTRANSFER RULES. We now have an understanding
of the operation of negative attraction, negative postposing, and
NEGCONCORD in four different dialects: Standard English (SE); the non-standard
white vernacular which does not transfer the negative to pre-verbal position,
of which New York City may be taken as an example (WNS1); the non-standard
white vernacular which does show NEGCONCORD to the verb, of which Atlanta
may be taken as an example (WNS2); and Black English (BE), which is relatively
uniform throughout the United States.
In ?5, we considered the possibility of combining these rules into a single
NEGTRANSFER rule. But ??6-8 show that there are two radically different kinds
of rules involved with different conditions, directions, functions, and frequencies.
Rather than use a single negative distribution rule like 16, we will first place
the negative in its pre-verbal position with the tense marker, by the normal
cyclical rule:
(139) NEGPLACEMENT: categorical
f 0 -V]
X - [+neg] - NP - e 1-be -Y
LLhave J
1 2 3 4 5 - 13 4+2 5
NEGATTRAC follows as a categorica
minate, with the single condition su
now stands for the complex of fe
+distributive, -fact -- (pred)]. The g
set of three superficially unconnected
formation is to be obligatory. In ?6
junction, depending upon the fact th
corresponding to the logical configu
hibited. But in terms of current wor
140 must still reflect the syntactic con
(140) NEGATTRAC II
X - INDET - [+neg]
1 2 3 -,13+2
Condition: Obligatory if 1 [ [
and 4, and 2 is [-stress].
It may now be possible to combine negative postposing and NEGCONCORD into a
single rightward transformation. But there is a great deal to be learned yet
38 The presumption in showing INDET directly before NEG in 140 is that the features of
Indeterminate are assigned to the entire noun phrase of the subject. Similarly, we see the
plural feature in English assigned to the entire NP containing the plural head noun. Thus
INDET in 140 extends over such phrases as any red-blooded American who wants to do his duty.
813
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
about the conditions governing negative postposing; on the whole, it is in complementary
social distribution with NEGCONCORD, but there seem to be some
cases where a speaker of any dialect will use negative postposing instead of
negative concord. Since we do not know the conditions under which speakers
will exercise this option, we will show the two transformations as separate rules
pending further investigation:
(141) NEGPOSTPOSING: variable
X - [+neg]- Y - INDET
1 2 3 4 -- 13 2+4
Conditions: (a) Y -[ INDET
(b) 4 cannot be dominated b
lexical NP which does not dominate 2.
(c) 4 cannot be dominated by an S adjoined to
THAT and commanded by 2.
(142) NEGCONCORD III: variable
X-[+neg]-Y- NDET
1 2 3 4 --1232+4
Conditions: (a) 4 cannot be dire
dominated by an S do
NP which does not dominate 2.
(b) 4 = vB in WNS2 and BE only; in WNS2, 2 and
4 must be clause mates.
The three NEGTRANSFER rules are now stated in reasonably simple forms. The
negative inversion rule would follow, which might be combined with other flipflop
rules. Negative raising also follows,39 so that we have
(143) I don't think the other guy had no chance either. (Danny W., 16,
Oscar Brothers)
(144) I told you, I DON'T believe there's no God. (Vaughn R., 16, Jets)
We will not explore further the conditions for negative inversion and negative
raising, since our main concern is with NEGCONCORD. Of all the NEGTRANSFER
rules, this shows the widest range of environment and dialect differentiation.
Condition (b) of 142 does not provide us with all the information we need to
characterize the uses of NEGCONCORD: this is best displayed in the matrix of
Table 4, which then replaces that condition.40
89 Either ordering of negative raising can be made to work here. If negative raising occurs
before NEGCONCORD, then the sub-case of NEGCONCORD governing transfer outside the
clause will apply; otherwise the major case applies within the clause. Since NEGCONCORD
seems to co-occur with negative raising with great consistency, the latter seems preferable;
but the issue is still open.
40 As we explore the relevant details of rules used by the speech community, it becomes
apparent that many rules will require a matrix of two or more dimensions to describe their
use. One dimension will be the various syntagmatic environments; the other will be the
range of dialects or idiolects which differ in their assignment of categorical, variable, or
prohibited status to a given environment. The ordering of the columns and rows of this
matrix for maximum scalability on a Guttman scale of implications is of considerable significance
in explaining the semantic, syntactic, or phonological motivation of the rule as it
814
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
TO INDETERMINATES TO THE VERB
WITHIN S OUTSIDE S WITIN S OUTSIDE S
SE 0 0 0 0
WNS1 X X 0 0
WNS2 X X X 0
BE 1 X X X
0 < X < 1
TABLE 4. Use of negative concord in four English dialects.
This matrix shows the gradual expansion of NEGCONCORD throug
vironments and four dialects as we move from the standard langua
The symbol 0 indicates that the rule is not used at all; X that it is us
and 1 that it is used regularly, that is, semi-categorically. Since NEG
is not used at all in SE, there is a row of zeros across the top. In WNS
in only two environments; in WNS2 in three; and in BE, in all four.
observe the semi-categorical use of NEGCONCORD to indeterminates
sentence, and a variable use in all three other environments.
Table 4 gives us the answer to the question asked in ?1: what proces
divergence can lead to completely opposing interpretations of ex. 1
no cat can't get in no coop? In this matrix we see the coupling of tw
(a) NEGCONCORD has lost its emphatic character in BE in the majo
ment, and (b) NEGCONCORD has spread into a new environment. We
the reinforcement of emphatic negation through a variety of gram
devices: the introduction of more quantifiers (She ain't in no seven
free-floating negatives (But not my physical structure can't walk through
negative inversion (Ain't nobody in the block go to school), and invo
concord with new quantifiers (Don't so many people do it). But the
our initial problem is in the pattern of Table 4: IT IS THE EXTENSIO
CONCORD TO VERBS IN FOLLOWING CLAUSES WHICH BRINGS ABOUT THE DIALECT
CONFLICT IN INTERPRETATION. White speakers are surprised by
they cannot envisage such a possibility. With the understanding
Table 4, our intuitions as speakers of other dialects begin to exten
ciation of the logic of ex. 1. There is no inherent reason why can't
a separate deep-structure negative of its own; it is simply the arb
tion of the dialects we know that NEGCONCORD does not operate
The extension of NEGCONCORD in ex. 1 does not lead to any seri
standing in ordinary conversation. Speakers of English are not un
sudden inversions of positive and negative. I couldn't care less seem
same as I could care less. In Maine and surrounding Eastern
young people say So don't I to mean the same thing as their par
Even without NEGCONCORD, we have learned that not every ne
what it might.
On the other hand, there is no reason to fear that BE speakers have lost their
develops in more or less favorable environments. The ordering of the linguistic environments
shows which linguistic features favor the rule, and the ordering of the non-linguistic
features shows either the point of origin of the rule for change in progress, or the social significance
of the rule for well-established variables.
815
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
grip on the distinction between positive and negative. The underlying semantic
features of quantifiers and negatives seem to be the same in all dialects of English,
which tend to differ in just the ways indicated here-by slight extensions
and adjustments of conditions on the same set of rules. Speakers of BE are just
as able as others to multiply or to cancel negatives, and just as able to operate
with two or more negatives in their deep structure. Consider the following discussion
from West Philadelphia:
(145) Huey: And he said, 'Nobody talks about my mother.'
Michael: Well I'm not nobody; I'm somebody. That's what he
said, 'I'm not nobody; I'm somebody.'
Games that play on the ambiguity of negatives, originating in deep or surface
structure, are popular among speakers of all dialects. In the same way, speakers
of BE play games with indefinites:
(146) Poochie: Hold it, Tokay. Somebody shot.
Michael: That's tough. 'Cause ain't nobody down here name some-
body.
Robby: You name somebody.
It is a reasonable assumption that speakers of all dialects of English have the
same range of linguistic and metalinguistic abilities, and that the most striking
differences in the surface forms of language represent the differential operation
of principles found in all dialects: different ordering of the same rules, specifications
and generalizations of conditions on a single rule, and different weightings
of conditions within the rule. A careful consideration of the obligatory and variable
properties of the rules involved will often throw a great deal of light upon
their function. We do not say that transformational rules have 'meaning' in
the strict sense of the term. But operations which shift, combine, delete, and
multiply have the effect of focusing, emphasizing, subfocusing, and eliminating
information. In the long run, most people find it necessary to emphasize, focus
on, conceal, or mitigate the same kinds of things. It is reasonable to assume that
speakers of any language will have a need for a wide range of more emphatic
and less emphatic negatives; if the range is constricted in one area it must be
expanded in another, and it is this kind of economy which we have been studying
in NEGCONCORD.
Though Table 4 has answered our initial question, we have not exhausted its
possibilities by any means. It is only a bare schema for NEGCONCORD, suggesting
how this rule may actually be located in time and space. This table provides a
grammatical model of the kind of matrices which Bailey has explored in phonological
rules. The matrix is a realization of weightings assigned to the variable
constraint on the rule. Such weights indicate the internal linguistic constraints
which govern the rule, and predict in what environments it will first become active.
As the rule develops, we frequently find that it becomes categorical in
the original environments before it operates variably in more extended, less
favorable environments. It is also possible for re-weighting of the environments
to take place.41 As Tables 2 and 3 indicate, the variable symbol X in Table 4 actu-
41 I.e., the relative order of the variable constraints can be altered. The direction of each
constraint is constant across dialects and age levels, but their relative strength can shift in
the course of time or age (see Bailey).
816
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
NEGATIVE ATTRACTION AND NEGATIVE CONCORD
ally stands for relations of 'greater than' and 'less than' in which e
greater than that above it and greater than that to the right. We h
begun a detailed examination of the use of NEGCONCORD, with the
of one dialect-BE. More detailed studies of the use of NEGCONCORD
of English dialects will allow us to organize the parameters of Tabl
greater confidence and deepen our understanding of the semantic fac
underlie the organization of grammars.
The most general implication of our investigation is in the contras
the categorical rule of NEGATTRAC and the variable rules of NEGC
While previous studies of dialect differences in rule systems have co
upon variability (Labov 1969, Garden, Bailey), here our primary focu
upon the categorical character of NEGATTRAC, as opposed to the v
semi-categorical treatment of NEGCONCORD in BE.
Rules such as NEGATTRAC which are uniform and invariant in all diale
language appear as resolutions of conflicts among abstract conceptu
of language. Whereas violations of many grammatical rules become
ceptable as they are repeated, violations constructed for such rules b
anything, less acceptable. They are truly uninterpretable: the listen
himself unable to construct a logical, social, or aesthetic rationale for
aly. We seem to encounter more than a contradiction in meaning; the
contradiction of intent. The listener is led on, and then perceives th
tention of the speaker was in an entirely different direction.
When a speaker begins with Anyone ..., we confidently expect som
general prediction; when he continues with ... doesn't ..., we are face
unaccountable reversal of polarity.42 There are practical jokes and p
rely on this mechanism of leading the listeners up (or is it down?) t
path. But a sentence like *Anyone doesn't sit there anymore is not
sensible, nor even in good taste; and if there is a joke, it is not on an
If we have correctly understood this phenomenon, our investigatio
ATTRAC and NEGCONCORD will have added considerably to our grasp
ture of linguistic rules.
REFERENCES
BAILEY, CHARLES-JAMES. 1971. Building rate into a dynamic theory of linguistic
tion. Working Papers in Linguistics, University of Hawaii, 2:9.161-34 [Rev
Variation and language theory, 1971.]
BAKER, C. L. 1970. Double negatives. Linguistic Inquiry 1.169-86.
BELLUGI, URSULA. 1967. The acquisition of negation. Harvard University dissert
CARDEN, Guy. 1970. A note on conflicting idiolects. Linguistic Inquiry 1.281-90.
FILTMORE, CHARLES J. 1967. On the syntax of preverbs. Glossa 1.91-125.
42 This effect of leading in two directions at once can be observed in some oth
structed violations of categorical rules of discourse. Some of the same uninterp
character clings to such violations as *Hello, I'm Bill Labov. How do you get to the
State Building? Introductions are anomalous if the speech event is 'asking for dir
But, as Erving Goffman has pointed out (personal communication), this violation
striking than *Joan of Arc is dead. How do you get to the Empire State Building? T
is simply uninterpretable nonsense. The strong reaction against the introduction is
begins with an appropriate series of moves for one type of speech event, then s
switches to another.
817
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms
LANGUAGE, VOLUME 48, NUMBER 4 (1972)
JESPERSEN, OTTO. 1924. Philosophy of grammar. London: Allen & Unwin.
KARTTUNEN, LAURI. 1971. Some observations on factivity. Papers in Linguistics 4.55-70.
KEENAN, EDWARD. 1971. Quantifier structures in English. Foundations of Linguistics
7.255-84.
KIPARSKY, PAUL, AND CAROL KIPARSKY. 1968. Fact. Recent advances in linguistics, ed.
by M. Bierwisch and K. E. Heidolph, 143-73. The Hague: Mouton.
KLIMA, EDWARD S. 1964. Negation in English. The structure of language, ed. by J. A.
Fodor and J. J. Katz, 246-323. Englewood Cliffs, N.J.: Prentice-Hall.
LABOV, WILLIAM. 1966. The social stratification of English in New York City. Washington,
D.C.: Center for Applied Linguistics.
-. 1969. Contraction, deletion, and inherent variability of the English copula. Lg.
45.715-62.
- . 1970. The study of language in its social context. Studium Generale 23.30-87.
-. 1971a. The place of linguistic research in American society. Linguistics in the 1970's
(pre-publication edition), 41-70. Washington, D.C.: Center for Applied Linguistics.
-. 1971b. On the adequacy of natural languages, I: the development of tense. Mimeo-
graphed.
-- . 1971c. Methodology. A survey of linguistic science, ed. by W. O. Dingwall, 412-97.
College Park: University of Maryland.
; PAUL COHEN; CLARENCE ROBINS; and JOHN LEWIS. 1968. A study of the nonstandard
English of Negro and Puerto Rican speakers, I-II. (Co6perative research
report 3288.) New York: Columbia University. [Reprinted by U.S. Regional Survey,
3812 Walnut Street, Eisenlohr Hall 202, Philadelphia, Pa.]
LAFFAL, JULIUS. 1965. Pathological and normal language. New York: Atherton Press.
LAKOFF, GEORGE. 1971. Introduction to linguistics and natural logic. University of
Michigan, mimeographed.
LAKOFF, ROBIN. 1969. Some reasons why there can't be any some-any rule. Lg. 45.608-15.
LANGACKER, RONALD W. 1969. On pronominalization and the chain of command. Modern
studies in English, ed. by D. A. Reibel and S. A. Schane, 160-86. Englewood Cliffs,
N.J.: Prentice-Hall.
REICHENBACH, HANS. 1947. Elements of symbolic logic. New York: Macmillan.
Ross, JOHN R. 1967. Constraints on variables in syntax. MIT dissertation.
SAVIN, HARRIS. 1971. Every any means every. To appear in Proceedings of the Conference
on Current Problems in Psycholinguistics, ed. by J. Mehler and A. Bresson.
Paris: Centre Nationale de la Recherche Scientifique.
SHUY, ROGER; WALT WOLFRAM; and WILLIAM K. RILEY. 1967. A study of social dialects
in Detroit. (Final Report, Project 6-1347.) Washington, D.C.: Office of Education.
STOCKWELL, ROBERT; PAUL SCHACHTER; and BARBARA HALL PARTEE. 1968. Integration
of transformational theories of English syntax. Los Angeles: University of California.
WEINREICH, URIEL. 1966. Explorations in semantic theory. Current trends in linguistics,
ed. by T. A. Sebeok, 3.395-477. The Hague: Mouton.
WOLFRAM, WALT. 1969. Linguistic correlates of social stratification in the speech of
Detroit Negroes. Washington, D.C.: Center for Applied Linguistics.
[Received 25 October 1971]
818
This content downloaded from
147.251.98.131 on Wed, 16 Feb 2022 14:19:44 UTC
All use subject to https://about.jstor.org/terms