PORT RESUMES
ED 010 871
AL.000 021
THE LINGUISTIC VARIABLE AS A STRUCTURAL. UNIT.
BY- LABOV, WILLIAM
PUB DATE
ECRS PRXE !4F -$0.59 HC -$0.64
21P.
65
CESCRIPTCRS- *ENGLISH,- *SOCIOLINGUISTICS, *PHONOLOGY,
*NONSTANDARD DIALECTS, SOCIAL FACTORS, ETHNIC GROUPS,
LANGUAGE STYLES, ICIOLECTS, URBAN LANGUAGE, DIALECT STUDIES,
*LINGUISTIC THEORY, MEASUREMENT, NEW YORK CITY
RESEARCH ON ENGLISH PHONOLOGY IN NEW YORK CITY IS
DESCRIBED. CURRENT LINGUISTIC ,THEORY IS CONSIDERED UNABLE TO
ACCOUNT FOR MASSIVE "FREE VARIATION" IN THE PHONOLOGY OF THE
SPEECH OF THAT AREA. ISOLATED WERE PHONOLOGICAL VARIABLES
WHICH ARE ASSOCIATED WITH SOCIAL, STYLISTIC, ET(± -.1C, AND
INDIVIDUAL FACTORS IN NEW YORK CITY. QUANTITATIVE INDEXES
WERE CONSTRUCTED AND INTERVIEWING TECHNIQUES DEVISED TO
ISOLATE CONTEXTUAL STYLES. THE FINDINGS SHOW THAT THE SPEECH
OF NEW YORKERS IS HIGHLY DETERMINED BY CONTEXT AND
&OCICECONOMIC CLASS. THE RESEARCH PERMITTED ESTABLISHMENT OF
A STRUCTURAL UNIT, THE LINGUISTIC VARIABLE, WHICH IS A CLASS
OF VARIANTS SUCH AS MORPHS CR PHONES, 6HICH IS ORDERED ALONG
A CONTINUOUS DIMENSION, AND WHOSE POSITION IS DETERMINED BY
AN INDEPENDENT LINGUISTIC CR EXTRALINGUISTIC VARIABLE. THE
THEORETICAL CONSEQUENCE OF INTRODUCING THIS CONSTRUCT IS THE
ENLARGEMENT CF LINGUISTIC THEORY. THE AUTHOR SUGGESTS THE
ADDITION CF THE LINGUISTIC VARIABLE TO THE REPERTORY OF
STRUCTURAL LINGUISTICS TO DESCRIBE REGULARITIES IN LINGUISTIC
BEHAVIOR WHICH WOULD OTHERWISE REMAIN INACCESSIBLE. THIS
ARTICLE IS PUBLISHED IN THE "WASHINGTON LINGUISTICS REVIEW,"
VOLUME 3, SPRING 1966. (KL)
is
a
=
7.=-±Ne!i;Irr,crei ,r-r*7c'7rta.--;
akar/ A
4,1 tie 1221
.eir fa P.F.A01-1,ZDUCATION
AND WtLPARE
(Office of Education
1:17/ g&qtyriet tias-$tean
reprociamd 6xactly
.11tifitin Or-orgon..:z:: ix;gnating it. Points as received from Ws
stated dr, pot n.?-Lizu-;.atiy
view or opinlom
rearesent official of
posirtori or II
wilco of-Education
II 8 0 000 Ti
clINGUISTIC VARIABLE AS A STRUCTURAL UNIT
e-4
William Labov
Columbia University
en-
I
=
AL 000 021
1.
Introduction:
the implication of sociolinguistic research for
theory
For much the greater part of the twentieth .cen.ury,
general
linguistic
linguistics
has been concerned with a very general type of lang% le, independent
particular speaker, the listener, or the speech community in which
of
the
language
r
occurs.
Considerable progress has been made by adhering to this abstract conception of language--in phonemics and morphology, and now more recently in syntax. Yet in the last few years, there has been a growing interest in the study
of language in the context of the speech community.
study as 'sociolinguistics',
Some
conceive
of
this
and envisage a new subdiscipline of linguistics,
with its own subject matter and its own problems.
Indeed, there are important
problems of language planning and multilingual conflict
which require such an
approach. However, there is a completely different approach to sociolinguistic
research, in which these studies are considered as an essential part of general
linguistics.
This is the point of view that will be prese.Led
phis aLscus-
sion, There is good reason to believe that many problems of general linguistic
theory that have been argued to a standstill for decades, can be solved by extending the range of data to be considered.
show how critical
In this report,
I would like to
problems
of cross-structural variation can be reanalyzed
and solved by the introduction of quantitative data from empirical research in
the speech community.
It is the introduction of this data, and the new regularities that appear
within it, which has strained our general linguistic theory beyond its limit.
Problems which
have
'wag
been recognized in the margins and interstices of
linguistic theory are suddenly thrown into high relief.
The fact that /ruwt/
and /rut/ are the same words, and yet somehow different words; the massive intersection of schwa with all other English vowels: the overwhelming problem of
accounting for gradital change from one categorical structure to another--these
problems all depend upon regUlar alternations which have no place in our
gen-
erill linguistic theory,. and their importance is suddenly magnified by the many
new and unaccountable regularities thatare found in sociolinguistic research.
4
r.
IV
ceetimeiciesszsczze=remeoe-=t7
e-eeeeeeeeeee
Labov
2.
"free variation" in New York. City phonolny
The immediate problem: massive
from a very
problem which will be confronted here stem's
studies of the New York City speech
particular problem exemplified in the many
phonetic 4.etail,
comiunity.1 These earlier studies provide -keen observation of
phonemic system as well but they are weakest
analses
of
the
and thoughtful
which is characteristic of New York City.
in treating the widespread variation
read that New Yorkers show "free variation"--in
For one item after another, we
/r/, in the vowel of bad, ask, dance, etc.,
the use of final and preconsonantal
daily, etc., and many others. For example,
in the glide of the vowel in sail,
of /r/ in New York City:2
Allen Hubbell says this about the use
The
a
theoretical
number of New Yorkers exThe pronunciation of a very large
might most accurately
hibits a pattern in these words that
Such
of any pattern.
be described as the complete ansence
before a consonant or a
speakers sometimes pronounce In
in a thoroughly haphazard
pause and sometimes omit it,
fashion.
.
of the speech system of New Yorkers
Is it possible that such a large part
The idea goes against the grain of our conis the product of chance factors?
(
structured type of human behavior. Free
ception of language as the most highly
in the
sense of irreducible fluctuations
variation certainly exists, in the
3
condiuioning,facLor.
si6nificani.
sounds of a Iarvaage wichouc any one
University of Michigan
Yakira A. Frank, "The Speech of Neu York City",
Pronunciation of English in New York
dissertations 1948; Allan F. Hubbell, The
1950); Hans Kurath and R.I. McDavid, Jr.,
City (New York: Ring's Crown Press,
Atlantic States (Ann Arbor: University of
The Pronunciation of En jish in the
"Let's Take Another Look at New
Michigan Press, 1961); Arthur J. Bronstein,
37:13-26, 1962.
York City Speech", American Seeech
1
Li
F
2a
cit., p. 48.
economic, medical,
3 /f ...in a very wide range of fields, including social,
fluctuation or
statiotics...there is a characteristic irregular
and scientific
the
events, and phenomena,
variation in the behavior of individual objects,
within the context under discussion."
details of which are not predictable
in Modern Physics (New York: Harper, 1961),
David Bohm, aegalty/ and Chance
this sort from significant
Bohm distinguishes chance contingencies of
p. 22.
the
empirical study: bcth are evidence of
causes which can be isolated by
but differ in the number and relative promiregularity of natural processes,
The characteristic step in the misapplication
nence of theecauses involved,
is to assume the absence of significant
linguistics
in
"free
variation"
of
empirical investigation.
causes or conditions without
5
a
irk.
I
1-,"71:77
I
Washington Linguistic lestia0
of entire phonemic
.
this is quite a:-differeAtliatt!r: _iteondens,osciLlations
and disappears as a whole;
the,sec_-of inglOing_phonemes appears
..categoriesr-
continuous
membership; one5rmnd class shows a
large word classes shift phonemic
The problem,- then, might be
boundaries.
phonemic
/three
transition across
factorsikich, govern the occurrence of in
what
are
the
stated in this way:
and having isolated such factors, we
in
New
York
City?
and other variables
be integrated into the phonehow -can these new structural elements
-must ask:
mic system of the speech community?
t
3.
linguistic performanCe
Methods: for the quantitative study of
is the
of speech behavior in New York City
for
the
complexity
One reason
ethnic
dimensions: social, stylistic,
takes
place
on
many
fact that variation
A number of new
commingled in the final result.
factors
are
and individual
These.have been
this variation.
methods were required to analyze and measure
4 here the steps may be summarized in enough deelsewhere;
described in detail
regularities to be discussed were actually
the
linguistic
show
how
to
tail
required.
derived, and what new conceptual approaches were
carried out in exploratory interthis
procedure,
Tice
first
step
in
a.
nf tho elvfor variables, rhat carry social sienificance.
viawel 4t1 0,0
because of their high frequency, their
Phonological variables are preferred,
codability, and wide distribution throughimmunity to total suppression, their
interviews suggested the existence
Seventy
exploratory
population.
out this
of
*stylistic and linguistic patterns for many
of corr&lations between social,
111111.41.11.11,AM.
1111111.11100.1110.110.0".11001111
Englishin New York City", (SSEN), Columbia
'The Social Stratification of
"Phonological Correlates of Social StratificaUniversity dissertation, 1964;
Hymes (eds.) The Ethncgraphx of Communication
tion" in John J. Gumperz and Dell
of a Linguistic Change in Progress", paper
(to appear); "Subjective Dimensions
"The
of America, Chicago, December 1963;
Linguistic
Society
-delivered before
Structures", paper delivered beReflectior of Social Processes in Linguistic
Boston, April 1964, to appear in Joshua
fore Eastern Sociological Society,
"Rypercorrection by the Lower
FishmanIed.), A Reader in ack211plujItica;
Symposium on
in
Middle Class as a Factor in Linguistic Evolution",
Isolation of the phonological variaSociolinguistics, Los Angeles, May i964.
of a Sound Change", Word
bles was first discussed in "The Social Motivation
4
19:273-309, 1963..=
6
41.
IttrOOVAIrillidStflzteefraiNMIMICZedicessdzkoriza.6110:61111
Lebos/.
examples of "free variation" by earlier
'items
-that
hadAieditagged
as
11.6
larger numvariables were established, and a
workers, Five Main phonological
variables which seemed to show significant
ber of.phonological and grammatical
5
(r), the
are:
main
variables
five
The
patterns of social distribution.
(eh), the height. of the vowel in
/xi;
occurrence of final and preconsonantal
Paul, office,
(oh), the height of the vowel in
bad, ask, dance, laugh, eic.;
the first confricative, affricate or stop as
(th),
the
use
of
talk, etc.;
variable
and (tilt), the corresponding voiced
sonant of thing, thoght, etc.;
in this
then, etc.
for the
step is the construction of quantitative indexes
taking into account every occurrence of
exact measurement of the variables,
each phone on a
The analyst codes
defined.
the
word
class
the variable in
in the case of (r) to five
simple-numericaf scale, ranging from two categories
of these ratings
categories in the case of (eh) and (oh). A numerical average
variants and the distribution from
the
range
of
the
index;
is the basis for
stalater in various measures of linguistic
the median variant is considered
b.
The
second
bility and deviation.
This sample must repreThe third step is the selection of a sample.
success
so that one can estimate the relative
sent a well-defined population,
the sources of error, and at some later"
in studying this population, estimate
what changes have taken place. In a
population
to
determine
time restudy the
alone is a difficult task; the
large city, the enumeration of a census tract
is to graft the linguistic study
Most efficient way to approach this problem
and conduct a secondary survey of a poron to an earlier sociological study,
concentration on linguistion of the earlier sample. This allows themaximum
c.
in reliability.
tic probleMs, with the greatest gain
study of New York
The main base for the descriptione the sociolinguistic
using as'a base.ehelsample
a
1961. Thii .Aelettion was
City English was a survey of the Lower East Side,
drawn by the Mobilization for Youth analysts in
in turn the' adult
stratified' random sample, from which lit selected
native
parentheses' symbOlize
OhOnernie'ettegoty4 (r) ia the linguistic
iifiguoic V4ialtieer thist Iii is the
and (r)-35
5
the notation to be used
X..-
throughout "this. papers
ia,
particular value or variant of the variable,
liariWte, (t4) is a
the variable.
is an AVerage'itidex adort for
7
-
Imimmimumm-
t4.
'Washington ?Atavistic Review
with repspeakers--of-Uglish who hatitiot,itroved within the, previous Mose/ars,
resentation from all of the important ethnic groups,
econoMic classes.
age
levels
and
!nolo-
A-target sample of 195 adults was constructed for th4 lin--
guistic survey' and:ovet-a four month period in 1963,
terviewid --over SO per.cent--along
a total of 157 were in
with 56 children of
these
informants.
Mr. Michael Rad of Haverford. College joined me in this field work.
the- Inest-arttittl Steps-in this series of techniques'it the
isolation of contextual style!. One of the basic concerns in the constriction
of-an-interview form was an escape from the dilemma of formal context vs. repOnly through a series of formal interviews in the home
resentative sampling.
hope to acquire comparahle samples of behavior of a represantarive
d.
One
of
can we
I
1
cross-section of the population.
But we want to obtain not just one style of
the
speech--the whole range of stylistic conditioning is our object.. And in
style which is appropricontext of the formal interview, informants will uee a
it may appear that the subject is
On the face of the matter,
ately careful.
casual and relaxed; but we must assume that he has another style's more casual
This
form of speech, which he uses with his family and intimate friends.
assumption was fully justified in the tight of our final results.
In the approach used here, the range of contextual styles is conceived as
elicit co trolled samples of styles along
a continuum: by various devices,
samples which arc ordered by successively greater degrees of
this continuum,
In the
attention given to the speech prosess (that is, more audio-monitoring).
most informal style, casual speech (Style A), the minimum attention is focused
The range of successively more formal styles includes: Style B,
on speech.
careful splesb, the main bulk of the interview;
D,
the pronunciation of isolated words;
and
Style C, reader stvle; Style
Style D',
the pronunciation of
"Animal pairs, distinguished only by. the. variable in question.
The first
two thirds
eliciting casual speech.
of the interview is designed about the problem of
Two topics are embedded in the interview which allow
the subject to disregard the constraints of thejormal situation. One of these
used by
is a discussi,on of childhood rhymes, and the code of oral legislation
preadolescent. children. Here the, use of careful speech style is inappropriate,
A
5
Labov
-in so* cages actually impossible.
6
The second topic- is introduced by a ques.=
of being
:,;.3.04.: about :aituations2. in which. the informant- was in serious denser
As he. recaunts'such .4n-incidents -his-reinVOlvement in- the situation
ki lled.
dist_tractshis,at4ent4in from-the-interview situation, td we obtain spontaneems
speech,wiplviehe same phonological characteristics as casual speech, breaking
through the formal constraints of the situation.
Three other contextual situations. may be defined in which the constraints
interview
of the formal interview normally do not apply: speech outside- the
a
proper, speech to a third person. and long digressions not in response to
direct question.
In five contexts,
occur.
The actual occurrence
channel
cues
of
therefore, we find that casual. speech au
casual speech is defined by accompeniing
independent of the phonological variables;
pitch, volume; laughter; or heavy breathing.
changes
in
tempo,
If in one of the five contextual
situations designated, at least one of the five channel cues is recorded,
the
utterance- is marked as casuatmach, and the values of the phonological vari-
ables are used for the average index scores for casual speech.
The pattern of stylistic
To
this
illustrate
of
range
ariation
the way in which phonological behavior is correlated with
contextual
styles, we may take the example of one informant,
Susan Salto, 37 years old, a third generation New Yorker of Italian background,
who works as a sempstress and a beautician.
in the greatest part of the interview, registered as careful speech, Style
in 26 per cent
B, this informant used constricted /r/
variable
(0-2-6.8
(r)
was observed.?
In
this
case,
of the cases where the
the index is this percentage:
In the Section's marked casual speech, Style A,
she showed (r)-02.
"I won't go to Macy's any more, more, more/ There's a,
in
6F*r -example,
big fat policeman at the door, door, door/" it is almost impossible to repeat
these lines' using the /r/-pronouncing style of careful speech with low (oh)
vowels,
7The variable (r) includes all instances where orthographic or historical
.pr consonantal position. , except after mid central
I_occursi,in!.
vowels, as in her, were, wssis ,n-tese latter word classes are studied sepnrateiy
aSAiffOre4, variables'.
of the occurrences of the
All transcriptions are
constricted 10: was recorded.
The notatiat(046
means-'that in '26",per cent
ide ,..00w the ?tepe recordings of the interviews.
9
fte?temoific,
.
Washington Linguist:1.C- :Review--sheAtimped: toy (r)-58;
-
In--the more formaL:dired.tion.fStyle:-C,
where-..(r) was-the chief foods,
andjor Pr49,*
in Style 4._ to
(r)i;400; - In this
she consistently- dittinguished.sackfrompard'bSi-the
use,of constricted id is ._guar
equally regular progressions in the
Susan.Salto
showed
At the sage time,
he showed (eh)-20, a
In casual speech,
phonological
.variables.
other four
In
vowel-of bad was the same as that of bared.
inwhich
the
pronunciation
in reading
and more,-open vowels, with (eh)-28;
careful speech, she used more
In this most formal style, she ton--,
(eh) '40.
tyle, (eh)-36; in word lists,
half, cash, etc. with the same low vowel
sistently pronoulAced bad, ask, dance,
r:
f
as with bat.
In
general
terminology,
linguistic
the behavior of Susan Salto can be
phonemic categories:
described as an alternation between two
Id/
/r/
/eh/
/agil
have to
which we actually observed; we would
But to represent the regularities
perhaps with
transition between these categories,
the
regular
indicate somehow
a symbol such as this:
However,
the
/0/
/eh/
In
/mh/
arrows
shown
here have no place in the alphabet of
structural
the problem
have been used in the past to handle
Various
devices
terminology.
We might, for example, speak of coexistent
cross-Structural
variation.
of
not'only the two endBut the structural invariant here is
phonemic systems.
between them. We can speak of diaregular
transition
it
is
also
the
points:
but it can be shown that Susan Salto is not
lect mixtures in varying amounts:
transition is itself a characterborrowing from any one else's dialeCt. -This
If we should adopt the method of autoprescriptive
aic of ,her own
dialect,.
generative grammar for working with
iltrlarT:tidrebtly-iised with good results in
re-edit the text,
Wormant6-, -and .14.4 Susan Salto to
ltoge the i ,
then the categories /0/
together wi th 'the transitions.
Aiii4iffe4410PlicitY',-bu t an uninteresting
simplicity.
l0
;
We would
st
.
s.
.1
`ssatestlistsesegrarimegurievenewcznaragerage41662
'14507$0.501035MC'
1.2bov
5
,
style, as
we, might, ,,isolate _each,
eMployed7
here:
concept _of idiolect might:,be,
However, the over
attempt to describe. that.
and
a.sepaxate,diolect.,
idiolects are not= the most co-n3istent/
all: result of.,t4s,a,tuOy:
in.New-York.City. Om the contrary,
behavior
of
linguistic
and contramost explicable
appears as studded with oscillations
the speech of many individuals
all framework of
against
the
over
only when it is placed
dictions, and it is
that we can discern the
indicates that
unit
speech community
social and stylistic variation of the
that governs this behavior.
regular structural pattern
in variation.
different- type of problem
(oh)
provides.
a.
The variable
comparable with the sounds
by Susan Salto a.re
-.any single
of the variants used
other word groups,.
used in any
variant consis-indicating a mixture of
(oh)-25.,_
speech
is
_casual
[;tfti ], with a slower, less
centralized _form
Furthermore,
Her index score in
tently.
None
she does,not use
partly
before
that used by most Americans
different
from
rounded form W4], not much
she moves
but in reading style,
index
is
the
same,
Er]. In careful speech, the
reading 'a word
(oh)-29,
and
in
with
form
the lower
the
high,
overrounded,
list, to
more
towards
(oh)-33:
This last index indicates
few even lower vowels, approximating
the lower form OM with a
Eastern
back rounded vowel PO of
a mixture of
the low
New !England.
these variants are
of
function,
none
From the standpoint of cognitive
category.
within a single _phonemic
all
be
included
significant, and they might
with a wide
have ordered variation,
variation,
we
Instead of cross-structural
a category:
range of dispersion within
S
.
.
/oh/
.
'"
.
category. The
unit
as
an
in
think
of
such
a
We have been accustomed to
tokens within this
the.distribution
of
invariant,- ,.and
type /oh/ recurs as an
minimal -unit has no interior- Struc'that
is,
this
type cannot be sigriiiicints
structural regularity we have Obfor,
the
,like
to
account,
ture' n But W.wtitild
of the evolu',considering the structural analysis
serVed;;,fniterevimPortin-tiyi-iri
that (eh}
it will becdme increasingly evident
tia
ci
Itoili,Cify speech,
it
WeshirigtOn
Linguistic Review
categocannot be in*the phonemic
The
parallel
and (oh) are parallel units.
variation,.- with two units,
,
cross-structural
show
strucfor the front vowels
ries
We would like to add'a
Therefore
have only one.
where.the,backvowels
=
:
of contextual
followed in the isolation
the
procedure
behavAs the result of
in the phonological,
has
appeared
of regularity
from random- styles, a certain degree
(r), for example, is far
Her
use
of
situation in
ior of this one, informant.
Yet the linguistic
chance.
the product of pure
it is surely not
Almost every individual
regularities.
clarified by such
deal of /ri,
Mew YorkCity is not
some using a great
different
system:
with a
others
would seem to operate
(eh-1) variant exclusively,
the
high
dome using
In what wav can
some with very little;
between.
somewhere
in
and the majority
always using (eh-4),
for this speech community?
coherent
structure
there be said to be a
selection of informants and
importance of a systematic
It
At this point, the
becomes evident.
characteristics
of their social
that of social
the cross-classification
stylistic variation to
question
of
from the
is necessary to pass
found.
of regularity may be
nigher
degree
variation, where an even
5.
variation
-The pattern of social
be correlated with
siructure which might
of
social
There are many aspects
interunderlying regularities in
find
the
tn order to
linguiltic performance,
important elements
class is one of the most
Sor!io-economic
personal variation.
and correlation
urban communities,
in
complex
of social structure
. i Dingrelation.
lingu"is
es irrarfe
structurally
and
are
(eh) and
shows
that
of
9A greAt deal of evidence distribution in the population, direction
their"
patterns of 'subjective
functionally parallel:
several generations,
to
earlier, and is -subject
stylistic shift,.evolution.through
establtshed
in that (eh) was
and has not yet
response. They differ
(oh) is less prominent,
correction;
socio-economic
the most overt social
variable in the speech of all
significant
evidence
is shown
appeared as a socially
conclusive
12.The
most
8,
9,
11,
Ch. 7,
-classes._ See. SSEM covariation of these variables.
:`,n the Statistical
(oh)/
,...1
eiress=wiszetalstssesitmaPanc3sEIVSN027P--
Labov
used here is oGe developed by MobiliThe scale.of socio-economic ranking
indicators of
original data, using the objective
zation for Youth from:their
(of the breadwinner), education (of the
productive status based on occupation
10
These are combined
per capita).
family,
and
income
(of
the
breadwinner),
socio-economic class (SEC), rangwith equal weights into a composite-index of
ing from 0 in the lowest bracket to 9 in the highest.
data, a great many detailed correlaIn the analysis of the New York City
and the differences it; the Lehavior of the
tions for each variable were. found,
of the evolution of
five main variables were important in the reconstruction
11
the New York City vowel system.
Some variables showed fine stratification,
each of six subdivisions of the
and it was possible to show correlations witi
major groups.
Others showed sharp stratification into two
U-9 scale.
and divide the SEC scale into three
However, if we disregard the differences,
obtain the Very regitlar matrices of
equal sections for all five variables, we
basic
Table I.
for each
systematic variation for each variable
The 0-2 group may be considered
social class and for each contextual style.
and the 6-9 group
informally as lower class; the 3-5 group a$ Estina class;
12
These regularities show us that the speech of New Yorkers
as middle class.
socio-economic class; that the deteris highly determined by both context and
small as ten individuals,
mination is regular in its effects for groups as
Table
In
is
there
is
that the two effects are
five;
though not necessarily for groups smaller than
which indicates that social and stynot independent, but show a covariation
underlying process.
listic variation are results of a single
wwodwwwwwwwwwwwwwwwwwwwww,......worow.wwww.wwwwwww
of the selection.
10See SSEN Ch. 7, pp. 216-224, for a detailed discussion
of this index.
.
11
See SSEN Ch. 9, 12.
middle class 6-8, and
2A division of the middle class group into lower
of linguistic
2221E middle class, 9, was found to be a regular characteristic
and social behavior in all of the' more detailed studies.
1
13
Ifort=iree
eatnWOON.64.Wall
Washington -.Litisu.stf,-O -Review
.
TABLE 3.
.C1 1$s STRAri*TAIN OF THE. VARIAB
21,aarkgrot.....m0-2
(r)
02.5
(eh)
(oh)
23.0- 27.0
10.5
14.5
29.0 32.0
23.0 24.0 24.0
78.0 65.0 43.5
78.5 ,-.56.0
(th)
(dh)
23.5 49.5
21.0
14:
17
17
15
17
28 26
27 26
23 28 -26
15 28 26
22 28 26
27
27
27
26
21 30
23 . 30
27 .30
23 30
27 30
29
29
27
29
18
22. 14
13'
21
13
16
18
17
22
22
22
13
13
13
Class Group 3-5
(r)
(el..)
(oh)
(th)
(dh)
04.0
25.0
19.5
68.0
63.5
12.5 21.0 35.0
30.5 32.0.
, 28,0
.22.0 23.0- 24.0
53.5 27.0
44.5 34.0
55.0
ti
14:
26
21
auxisys2...21
(oh)
25.0 29.0' 55.5 70.0
27.0 30.0 -34.0 -35.0
20.0 .23.5 26.5 29.5
.(th)
25.5
(r)
(eh)
(dh),,
-12.5
16.5
10.0
16..5
13.0_
29
29
29
29
29
AIKAPPOINOPIIIMPOW11.01.10"Wiliq.b44".
....."1---
Labov
6.
rneqinguistiC -variable
behavior permits a more precide
The exact measurement of sociolinguistic
It
structural unit symbolized by the arrows above.
characterization of the
Whereas the linguistic variant is a
may be termed the linguistic variable.
class of variants which
particular item--a morph or a phonethe variable is a
is determined by
ordered along a continuous dimension and whose position
are
I
i
variable.
an independent linguistic or extra-linguistic
system of New York City, we
In the over all symbolization of the 1:Lanemic
This particular variable .appears as a frecan now insert the variable (r).
quency with which /r/ is observed in a givin environment.
.(r) = X /r/
In turn, X is correlated
where X is the frequency with lohieh /r/ appears.
class and contextual style:
with the social variables of socio-economic
X = f (S, C)
S = stylistic level
C = clas's status
More
precisely,
we can
make
some ?estimate of what the function f may-be.
model:
Table 1 does not depart seriously from the linear
f (S, C) = aS + bC + c
In order to specify the
One cannot proceed beyond this point at present.
to develop a numeri»
values of the constants a, b and c, it would be necessary
of style. Although Styles A
cal evaluation of the continuum of formality
the graph, it is evident that we can
through D' are shown equally spaced on
relations remain
only say that they are ordered qualitatively' quantitative
to be determined.
that some are
of the distribution of the variables show
correlatedwith age, reflectcorrelated with ethnic group membership; some are
With.the help of this infOring an underlying proCess of change in real time.
evolution of the New York City
matiOn, we can reconstruct the stages of the
variable permits us to
It was noted above that the linguistic
vowel sydtem.
and within categorical units such as the
describe systematic variation across
mechanism -for structural
It also allows us -to produce a plausible
phoneme.
Further
studies
change.
15
11.10, grACIf,
Washington Linguistic Review
=.Some linguistic variables
are
functions of 'other lingniitic. variables,
sysand are not directly influended by any factors outside of the liiiguistic
13
Others axe- functions-of- extra-linguistic factors: socio- economic
tem.
ethnic group,
-class,
age,
These extra-linguistic factors are in
and so-on:
increase,
constant process of change themselves:: social stratification may
children continue to react with and against
ethnic groups are assimilated,
reflects-this change by a gradual shift
their elders. The linguistic variable
for example, we may hear higher and higher
of distribution of its variants:
variants of (eh) in bad,
ask,
dance
in the course of several decades, until
the vowel passes from the status of a low front vowel to a high front vowel.
vowels appearing in some other
Identification of the principal variants with
phonemic
word classes may lead to a concentration of variants at a particular
level.
14
If the conditioning illictor disappears
completely,
the
linguistic
phonemic
variable loses its ordered structure, sad is resolved into one, or more
On the other hand, we can-see some phonemic categories gradually
categories.
in the case
entering into the systematic ordering of a linguistic variable:
as
in vivo,
of (oh), we are fortunate in being able to witness this process
working class and the
this item gradually-acquires social significance for the
lower class.
15
closely
The position of the LOW central vowel (ah), for example, is
While (oh) shows a significant didtribution according
correlated with (oh).
signifito style, class, age and ethnic membership, (ah) shows little social
the speech of a given percance of this type, and is essentially constant in
lax vowel
When (ah) does fluctuate, it is with replacement-of the short
son.
variable (r).
la/, and this is directly correlated with the use of /r/ in the
13
of varis the case with (eh), *where we find a heavy concentration
where, bared, and also
iants at the level-of (eh-2), the mid position of bear,
(eh-4), the low position of bat, bad.
14This
15
stylistic
The lower class informant's showed no consistent social or
differentiation of
variation for. (oh); working class informants showed some
styles; lower
the high form of casual _speech from slightly lower forms in other
variation from
middle class informants displayed a consistent and wide range of
with stylistic
the very highest to the very lowest variants, closely correlated
with the
Subjective response tests showed the lower riddie class
context.
greatest sensitivity to this variant, and lower class with the least.
16
.
't
,
linguistic variable as a structural unit
difficaty
in-the
use
of
ette
--One
structure of a New Yorkpr who never
is that we may-yrite (r) for_the phonemic
evidence to support this step, we
uses /r/ in his Onw':;i0e-i,A744 additional
reactions to the phonological
unconscious
subjective
the
study
of
may turn to
variables.
7.
1
1
11
di
.
-Subjective reaCcions to phonological _variables
significance of the phonological
our statements about the social
distribution in actual
inferences made from their
based
upon
are
variables
the New York
exactly whether the members_of
speech:. We would like to.know more
the
values of these variables and judge
do
react
to
the
speech
community
City
of the sociolinguistic structure
social status of the speaker on the basis
because subjective reacThe task is exceptionally difficult
just exhibited.
bilow the thre'shhold of conscious attentions to individual variables are well
mixed with reactions to
they
are
inextricably
situations,
In normal
tion.
what the speaker is saying, and
voice qualifiers intonation, articulation,
Side, a subjective
In the survey of the Lower East
general
appearance.
his
evaluations
successfully isolated unconscious
reaction test was developed which
So far,
-
I
variables.
of the informants to the five main
met, and
used in this subjective evaluation
16
The details of the methods
elsewhere.
have been reported
evidence for their reliability and validity,
that is, the
of (r)-positive response:
Here we may consider the distribution
characteristic of New Yorkers which is consispattern of subjective response
Tables
prestige feature.
recognition of /r/-pronunciation as a
tent. with the
thr actual use of In with -the
2 and 3 compare the distribution of
subjective
evaluation of /r/:
16
See SSEN -Ch. 11, and
Progress "_, cited. above.,
p.
Change in
Subjdctive Dimensions of a Linguistic
17
3
1
-
-
TABU 2-
VEN,AGE; -D= NIL
(r)
4,__,Bt SEC AND AGE
404.-
0 -1'
20-39
CO
40-
OD
.
00
'00
34
-06
09.
09
3
13
10
25:
'9
8.
suka-
'I, .1.
1-
TAB1E10.3
18
PERCENTAGES OF--(r)-POSIT-IVE RESPONSE "TO THE
TiO-CHOICE TEST BY SEC AND AGE
SEC
Asit
20-39
40-
0-1
275
6-8
100
100
-100
63
60
70
9
100'.
3
13
57
8
26
11
10
5
7
for all practical
great majority of New Yorkers are
speech (Style A). Only the younger Class,9 group.
/r/-less
in
everyday
purposes
college
avtiaucnt of /r / - -that is, the upper middle class,
Table 2
shows
that
the
shows any apprecis.ble
But in Table 3 the situation is radically
speakers, irrespective of class, shows
of
younger
The
entire
group
different.
show
the older speakers are quite erratic and
an (r)-pos=itive response; -while
educated group below tbe.age of 40.
.
no fiked pattern.
establishing the validity of
The subjective reaction test is important in
for' the New
structure which has been suggested
the over, all sociolinguistic
Z,would like to suggest that we consider the,
community.
Cityl,-sPeech
york
evaluative norms rather than by any
speech,communityas defined bye the set of
- universal
foetur4;pfimeesh performance.
144, 9t $SENp. 346.
2
S
!
P
4
7IR
c.-
-
;11==
-
aik.4:1"om
_
Labov
ay
in 'hit 'eve
-.fe11014- that as ind ividual who doe, not Use any hi
part-i0Patei-in_
ndor. pettqa in airy OCirrettatioti:
-OnVerkationi
the SOciolinguistge -6t703,40te:
of the
coiskinity as .a
whole
The Variable
particUllir vatue Of' the ftuAttioti
appeart in his phonological sytte4.
f (A, S, C)
4.
age
context
S
C = clime
for.this particular age, style, social status,
this
membership
indicates that the phoneme fr/
barring any further change in the, situation described
formality at
function, we can point to particular values for age or
will not appear.
by
Cr)
But,
which members of this social group will use the category
that the most systematic
Thus we have arrived at a very general result:
but that of the
structure As not that of the individual. for the idiolect),
general linguistic principle
speech community. This is it accord with our most
that lance ea_ is not an aspect of individual
as- enunciated by de Saussure:
behavior, but rather of social behavior.
S.
Conclusion:
the enlargement of linguistic theory
as a working tool,
The utility of the linguistic variable as a concept,
in this discusand as a structural element of analysis has been demonstrated
of linguistics where such
One might introduce examples from other{ areas
sion.
At this point, however, it would be
a unit may solve long-standing problems.
of introducing this element
useful to re- examine the theoretical consequences
into formal linguistic analysis.
discussed" actively
Behind al:: of the major linguistic' theories that are
assumptions about the nature of structoday, there seem to be a common set of
I would call the "categorical view". It
tural units. This set'of assumptions
'includes the assertions that all linguistic units are:
1.
'discrete
inva;lant.
3.., conjunctively defined
4.':.-44aiititiVoly different
.
:I.W.,414.141.4,w,is.1,11qeant.
'
.
ficiria:&tat,iithti by cleardut
that toe units are separated'
discOntinuities of 'form
or futxttorif
-1,0ears as precisely thksamo in each ,occurrence
19
despite the fact that'tokens
-t
...,
/
ugtidit4Y
to *aril;
defined,,, that there it
ft set
or properties isOcici-
with the unit which are in some 141y necessary, _essential,
as opposed to
otheiprOPii4es-WhiChati,accidental or redundant; by aualitative47 different,
at the units are completaty different from,_one another, and not distinguished
ordeied sequence.
is homogeneous eUMents it an
_A fifth, property, maylbe con-
although some structure' units are
aidered' as an extension of discreteness:
eo- mpoundid'of others, there is a limit to any such, subdivision,
and there, are
-
for which no fur-
for each field a set of atomic, primes, integral categories,
ther subdivision is possible.
_
These
appear
to
.
.
.
-,
-
properties
of linguistic categories are far from arbitrary., They
correspond well to the basic structure of
it everyday.
laUagelis
we deal. with
.
,
.
.
it is sometimes said that man is a.categorfzing animal:
it
is
eqUaliy appropriate to say that language is a categorizing. activity. Theaban.
donment'of an'y one of these properties' might be shown tohave.unfortunate. con-
sequences for linguistic analysis.
Nevertheless, some modification seems necessary,
types of variation Which seem to resist this strictly
AS we
have
seen,
cross-structural
ordered variatIon within a category
change
from one
set
for
there
are several
framework.
categorical
variation is one such type; the regular
is another;
the problems of dealing with
of categories to another has been discuskid.
Finally,
alternate
analyses
there is the problem that the categorical view often faces
which are equally consistent with the observed facts, and the
choice
between
the two cannot be made on the basis of empirical observation within the theory.
The
multiplicity
of methods
that have been developed for dealing with
Simple
variation seems to be an indication of the seriousness of the problem.
neglect" ,is one such. method,
perhaps based on the view that variation below- a
certain_ level, of structure is,automatically nonlinguistic.
AnOther approach
if, the Concept of the idiolect as ehe ultimate structural reality.
Coexistent
Syfte44_have been discussed:, The. common core and: over all pattetn
concept has beop
explored, although with serious results for the
concept of linguittic structure.
For dealing with the
problem of
variatien,,,linguistics,has4govideCaurgle of absplUte division bet
temporal
en
syn-
.chtenic .andJliachronic analysis,- 01ne-of Xhe iiOst_commonly used methods followed today is'that Of thy- school of generative ,grammar.
By setting aside the
WMmlb._ .
laboV
naVior
in -favor ,of_a normative, 44ited text,
it is
evil ettnet'4f-toptextuit
HOWeVer, it can
potisitzle.,a0-dispOlie,_04--great-deal of troublesome variation.
analysis have been
01104110i...pi the successes of this type of
out
be,,Joipte4.;
achieved_ -in
systematic--that is,
wher0,07:0at,ive_patterns. are.highly
in
Much further.
syntax--and this propeity#Ity not extend
here is to add a different type of unit
The solution that I anyadVocating
The
the linguLstic <variable.
linguistics:
to- the- repertory of structural
fact- -that it has an internal
variable ia.distinguishedfromiii;tategory'bythe
Variables are not primes, and they
divisible.
is'indefinitely
structure which
not be conjunctively deFurthermore, a variable may
do not contain primes.
its further extension may be governed
invariant,
in
so
far
as
fined. It is not
Thus'
linguistic or nonlinguistic.
variables,
correlations.withother
by its
/ih/ in New York City is the product
the gradual extension Of -(eh) to include
correlated.
social factors with which it was
of the gradual extension of the
In actual texts, we meet with
The variable is of course an abstraction.
basic step
themove from yariant to variable is the
However,
variants only.
indithe speech performance of the
which must be taken here. It implies that
trrough the assumption of an underlying
vidual or the groupis best explained
form, reform and dissolve. Empirical
linguistic continuum, in which categories
it is accomplished by showing
proof for this claim must be made in each case:
independent variable always produces
that the smallest possible change in some
a change in the dependent variable,
variable in quesr
that is, the linguistic
tion.
If X = f (S),
.then
.
X + 4 X = f (5 +AS)
of the continuously divisible charThe empirically established statement
for its use in
is a necessary condition
acter of. the lingUistic variable
with
(1) ordered, and (2) correlated
Variation
must
be
structural analysis.
conditions were not laid down, the
some independent variable. If such strict
provide an easy solution to almost any structural
...linguistic
variable would
OtobleM, but at the cost.of.eliminating all
advantages to be gained from introdUcing the
doubt remain to be
a Structural element, and some no
There are undoubtedly _many
lingUistic variable .as
method.
21