of SocialStratification

Correlates
Phonological
WILLIAMLABOV
ColumbiaUniversity
INTRODUCTION

AS WE approach the study of language in its social context, it seems that

by the very same steps we enter the study of small differencesin language
behavior. For many years, the structural analysis of sound systems has enjoyed,
and profited by, a kind of bold abstraction from such differences. Small differ-
ences within a system have been explained away as "free variation" or "social
variants," and we have concentrated on the abstract organization of constant
features. But to understand the dynamics of such systems, the mechanism of
their evolution, and their role in community life, it is useful to reverse this atti-
tude. If a feature of language is constant from place to place and speaker to
speaker, the fact will then appear to us pale and uninteresting. But the moment
we hear a differencebetween two speakers or two speeches, our interest is quick-
ened. Does the differencerecur? Is it generalizedin any context or social group?
Does it have social meaning? As we turn from the study of linguistic constants
to linguistic variables, we acquire more realistic methods of comparing systems
and measuringdifferencesbetween [Link], as we develop quantita-
tive methods, correlationsbetween linguistic patterns and other cultural patterns
begin to emerge.
The important word here is quantitative. As naive and native speakers of a
particular region and generation, we all receive a great deal of qualitative infor-
mation from small differencesin the speech of others. The linguist's task is to
construct quantitative measures by which such information becomes a precise
medium for comparison and further abstract manipulation.
There is an element of paradox in this emphasis on quantitative procedures.
One of the undisputed achievements of structural linguistics was the analysis
of various continuous dimensions into qualitative units, conceived as absolutely
different from one another, bounded by sudden transitions in terms of binary
choices. This model was a successful representation of a synchronic structure
with only one function: that of cognitive communication. Its limitations have
become increasingly apparent as linguists begin to analyze the many functions
of language in the context of the speech community (Hymes 1962.). Further
questions are raised by the search for the origins and dynamics of linguistic
change: in several empirical studies it has been possible to outline a detailed
quantitative structure which lies below the level of the quantitative functional
unit.
In a previous study of the speech of the island of Martha's Vineyard (La-
bov 1963), a numerical index was constructed to measure a complex distribu-
tion of small differences in the diphthongs /ai/ and /au/. The linguistic pat-
164
 Phonological Correlates of Social Stratification 165
terns were then correlatedwith patterns of social forces, and from this juxtaposi-
tion there emergeda unified social motivation for the sound change under study.
As this line of attack was pursued in the urban core of New York City, a
pattern of social stratification was encountered which cut across all other forms
of social differentiation. One would expect that this pervasive social pattern
would be reflected in a number of linguistic variables; the social significance
of these variables for the members of the speech community would be indicated
by correlations with objective indicators of class stratification. In this case, the
analysis of social variation in language was not an adjunct to the analysis of
the phonological system of functional units, but rather a necessary preliminary
to an adequate analysis. In New York City, the speech of most individuals
shows a great many oscillations and fluctuations, seemingly in defiance of the
need for a coherent linguistic system for rational communication; but when
this behavior is placed in the context of the structure of stylistic and social vari-
ation characteristicof the community, it appears as part of a highly determined
system.
THE SURVEY OF THE LOWEREAST SIDE
The data for this discussion will be drawn from a completed study of the
social stratification of English in New York City (Labov 1964), and particu-
larly a survey of the Lower East Side conducted by the author with the assist-
ance of Mr. Michael Kac of Haverford College. This section of New York City
was the focus of a sociological survey undertakenin 1961 as a preliminaryto the
Mobilization for Youth Program. Through the cooperation of Mobilization for
Youth and the New York School of Social Work, it was possible to use the sam-
ple already constructed as a base for the linguistic survey.? The linguists were
thus relieved of the difficult tasks involved in the enumeration of 33,000 dwell-
ing units, the construction of a random sample of 100,000 residents, the deter-
mination of comparative social status, and the exploration of social aspirations
and attitudes. Furthermore,our secondary survey had a great advantage in effi-
ciency, in that we were able to construct a stratified sample of only those sec-
tions of the population with which we were immediately concerned-the adult
native speakers of English who had lived in the area two years or more. Since
the socio-economic, racial, and ethnic composition of the population was al-
ready known in detail, it was possible to draw the minimum number of subjects
from each subgroup which would yield an adequate characterizationof the lin-
guistic habits peculiar to that subgroup.2
The relative socio-economic status of the individuals who were interviewed
had been determined from the completed sociological survey by an objective
index of productive characteristics constructed by Mobilization for Youth
(Michael 1962). This index combined three indicators: occupation [of the
breadwinner], education [of the informant3], and a family income figure. De-
tailed correlations of the individual indicators with linguistic behavior showed
that no single indicator was as closely correlatedwith linguistic variables as the
166 The Ethnography of Communication
combined index of three indicators.' This finding lends confirmationto the lin-
ear nature of the socio-economic index, and suggests that the abstraction of
class status is indeed required for an adequate analysis of social and linguistic
behavior.
The socio-economic index divided the population into ten classes, from 0
to 9. Linguistic and social patterns agreed generally in the division of this scale
into four class groups, which may be informally labelled: 0-1, lower class; 2-5,
working class; 6-8, lower middle class; 9, upper middle class. (Classes 2 and 5
are marginal groups, sometimes following the behavior of the lower and lower
middle class respectively.)
The target sample for the linguistic survey consisted of 195 adult native
speakers of English who had remainedin the Lower East Side from 1961 to 1963.
Eighty-one per cent of this target sample was reached for information on lin-
guistic behavior. Complete linguistic interviews, in tape recordedsessions lasting
from an hour to an hour and a half, were conducted with 122 informants. The
data to be given here represent the speech of 81 informants who were raised in
New York City;5 the data for the one-third of the sample who were not native
New Yorkers provide a valuable base for distinguishing those linguistic struc-
tures which are peculiar to New York City from those characteristicof the East-
ern United States as a whole, but this material will not be presentedhere.
SELECTIONOF THE LINGUISTICVARIABLES
There are four principal criteria which are used in selecting a linguistic vari-
able for quantitative study. For the greatest utility in an investigation of this
type, the variable should be high in frequency; have a certain immunity from
conscious suppression;6 be an integral unit of a larger structure; and be easily
quantified on a linear scale. All of these criteria point to segmental phonological
features as the most useful, and we rely upon them for our main quantitative
work.7
Five main phonological variables were used in the survey of the Lower East
Side. In the following discussion, a phonological variable will be indicated in
parentheses, e.g., (r), as opposed to the phonemic unit /r/ or the phonetic unit
[r]. The individual values of the phonological variable may cover a wide range
of phonemic and phonetic units: they are indicated by figures within the pa-
rentheses, e.g., (r-1). Average index scores for an individual or a group are shown
by figuresoutside the parentheses,e. g., (r)-22.
The five phonological variables are: (r), registering the presence or absence
of final and preconsonantal/r/; (eh), indicating the height of the vowel in the
word class of bad, ask, half, dance [as opposed to bat, back lap]; (oh), indi-
cating the height of the vowel in the word class of off, chocolate, all; (th), the
realization of the first consonant of thing, thought, three, as stop, affricate, or
fricative; and (dh), a correspondingindex for the voiced initial consonant of
then, this, the, etc. The discussion to be presented here will concern (r), (eh),
and (th).
 Phonological Correlates of Social Stratification 167

THE ISOLATIONOF CONTEXTUALSTYLES

The phonological variables listed above are especially useful because they
show both interpersonaland intrapersonalvariation. In the speech of the great
majority of New Yorkers, these variables follow a pattern of continuous and reg-
ular variation through different styles and contexts. This is a major opportunity
and yet a major problem. We can profit from this variation to learn more about
the structure of language than we would in other regions, like Martha's Vine-
yard, where most speakers have only one style. But first we must control the
context and define the styles of speech which occurwithin these contexts.
To study social stratification, we need random sampling. To complete ran-
dom sampling, we need structured, formal interviews. Yet formal interviews de-
fine a speech context in which normally only one speaking style occurs: careful
speech. The bulk of the informant's speech production at other times may be
quite different. We can hear this casual speech on the streets, in restaurants,at
Coney Island. Yet anonymous observations in these contexts will also be bi-
ased. These are special groups-that hang around street corners, go to Coney
Island, or talk loud enough in public places to be overheard. The interview re-
mains an indispensabletool for an accurate cross section of the entire population.
What then can be accomplishedwithin the bounds of the interview?
First, we can measure linguistic range by shifting context and style from
careful speech in the direction of more formal behavior. We give the informant
standard texts in which the variables are concentrated, and record his reading
style. Next, we give him lists of isolated words, which follow patterns of pho-
netic alternation, or ask him to pronounce consciously certain minimal pairs
which he has just pronounced unconsciously in the reading. This procedure
yields two additional speech styles of increasing formality. But we cannot rest
content until we somehow open a window on that every-day speech in which
the citizen scolds his children, jokes with his friends; and orders a slice of apple
pie. It is the familiar problem of whether the light is on or off when the refrigera-
tor door is closed.
To solve this problem, we must construct interview situations in which cas-
ual speech will find a place, or permit spontaneous speech to emerge, and then
set up a formal method for defining occurrencesof these styles. By casual speech
I mean the normal speech used in informal situations, where no attention is di-
rected to language; by spontaneous speech, I mean a pattern used in excited,
emotionally charged speech when the constraints of a formal situation are dis-
charged. Both of these styles share the feature of a reduction in audio-monitor-
ing so that speakers are less able to preserve an acquired prestige pattern at the
expense of an earlier, motor-controllednative sound production. In our results,
spontaneous and casual patterns are equivalent, and they will be discussed
jointly here as casual speech.
We acknowledge the presence of casual speech when at least one of five
contextual situations prevail, and also at least one of five nonphonologicalchan-
168 The Ethnography of Communication
nel cues. The contextual situations are: speech outside the formal interview,
discussion with a third person, remarksor monologues not in direct response to
questions, and two topics within the interview itself. One of these topics is a
discussionof childhood rhymes and customs; spontaneousspeech is normalin
My mother,your motherlives acrossthe way,
Two-fourteenEast Broadway,
Every night they have a fight
And this is what they say ...
A later section of the interview leads up to the question, "Have you ever been
in a situation where you thought you were in serious danger of being killed,
where you thought, 'this is it'?" If the answer is "yes," we ask, "What hap-
pened?" There is a psychological pressure to prove that this was a real, not an
imagined danger, and the speaker often becomes involved in his narrative to the
extent that his attention is entirely focused on this re-enactmentof the past.
When speech occurs in one of these five situations, we look for one of five
channel cues: changes in tempo, pitch, or volume; laughter; or changes in
breathing. If at least one of these occurs, the variables in this section are tabu-
lated under casual speech.
In tabulating even a few interviews, it becomes apparent that these formal
definitions correspond to some regular patterns of alternation characteristic of
New Yorkers. One middle class speaker, for example, uses 19 per cent /r/ in final
and preconsonantalposition. In casual speech, his usage drops to 00 per cent.
In the other direction, he increases to 24 per cent in reading style, and jumps to
53 per cent /r/ in isolated word lists. An upper middle class speaker may follow
the same pattern at a higher level, or a working class speaker at a lower level.
For example:

(r)
Casual Careful Reading Word
3 Individuals speech speech style lists
Uppermiddleclass 69 85 96 100
Middle class 00 19 24 53
Workingclass 00 05 14 29

Almost any small section of our population shows this typical two-way stratifi-
cation, following a hierarchy of styles and a hierarchy of social status. One can
therefore turn to the data from the group of 81 speakers with every expectation
that it represents the habits of the population as a whole.
CLASS STRATIFICATIONOF (th)
The first variable to be consideredhere is (th), the realization of the voice-
less initial consonant in thing, through, etc. This variable is not at all peculiar
to New York City: its social significance is roughly the same in most areas of
the United States. A phonological index is used to codify this variable using
three particular values: (th-1), the fricative form [O]; (th-2), the affricate
[tO]; and (th-3), the lenis stop [t]. If the informant uses only the fricative
 Phonological Correlates of Social Stratification 169
form (th-1) in a particular stylistic context, his average index score for that
context is (th)-00. One point is assigned for each occurrence of (th-2), and
two points for each occurrenceof (th-3); the total number of points, divided by
the total number of occurrences of the variable, and multiplied by 100 yields
the (th) index score. Thus consistent use of the stop form would be assigned
a score of (th)-200. No native New Yorker uses only stops or affricates: this
feature is a true variable for all but a few informantswho used only the prestige
form.
Figure 1 shows the class stratification of (th), for the 81 native New York
informants, in four contextual styles. The vertical axis represents average (th)
index scores. Along the horizontal axis are shown contextual styles, from the
most informal at the left to the most formal at the right. At A are the values for
casual speech; B, careful speech; C, reading style; and D, the pronunciationof
isolated words. The average index scores for each of five class groups are plotted
for each contextual style, and the values for each class group connected along
straight lines. At the bottom of the diagram, the upper middle class shows av-

100 Socio-economic
class groups
SEC 0-1 lowerclass
2-4 workingclass
7-6 lowermiddle
class
9 uppermiddleclass
80

(SEC
o 0-I
60

2-4

40

20- 5-6

?A B C D
casual careful reading word
speech speech style lists
CONTEXTUALSTYLE
FIG.1. Classstratificationdiagramfor (th).
 170 The Ethnography of Communication

erage scores very close to zero in all four styles: this group departs very little
from the prestige standard used by radio and television [Link] second-
ranking status group, the lower middle class, is shown here divided into two
halves which use the same amount of stops and affricates in everyday speech,
but diverge in their ability to approximatethe prestige standard in more careful
speech. All of the middle class subgroupsare divided by a wide margin from the
usage shown by the two lower classes. The working class group, representing
mainly employed manual workers, shows a (th) score of 70 in casual speech,
and shows a marked decline only in reading style and isolated words. The high-
est values of all are shown by the lowest ranking class group; the lower class is
a very mixed set of individuals, many permanently unemployed, from broken
homes, with little education.
The pattern shown by this diagram we may call sharp stratification, in that
the population is divided into two radically different sections by their use of the
(th) variable. A similar pattern may be derived from a tabulation of the voiced
counterpart, (dh). The regularity of the pattern is a striking feature: the de-
tailed class stratification found in casual speech is repeated and confirmed by
the values found in the three successively more formal styles. Similarly, the pat-
tern of stylistic stratification followed by the upper middle class is repeated and
confirmedby the stylistic stratification followed by four other class groups. This
array of regular relations has a self-confirming,self-governingproperty which is
characteristic of linguistic systems where "tout se tient." In a regular structure
of this type, any given datum might be replaced if missing within narrow lim-
its, since its value is determined by the values of all neighboring points. It is
difficult to conceive of any nonlinguistic cultural measure which would provide
an equally complex set of quantitative relations with other forms of social be-
havior.

CLASSSTRATIFICATION OF (r)
The next variable to be considered is of a radically different type. Whereas
(th) is a linguistic variable with a stable social history, common to large parts
of the United States, the variable (r) is involved in a rapid linguistic change
which is especially characteristic of New York City. The traditional pattern of
New York City speech, as described by Babbitt (1896), Frank (1948), Hub-
bell (1950), and Kurath and McDavid (1961), was consistently "r-less": that is,
the phoneme /r/ did not appear in final or preconsonantalposition. In this pat-
tern, god is homonymous with guard, sauce with source, bad with bared. The
prestige pronunciation which was superimposed upon this pattern was heavily
influenced by Eastern New England and British speech, and was also r-less. In
recent decades, a new prestige pattern has been superimposed upon the speech
of the city, based upon an r-pronouncingdialect characteristicof other Northern
regions outside of Eastern New England. This has replaced the earlier prestige
pattern almost completely in the speech of our informants. Figure 2 shows the
class stratification of (r) which has resulted from the importation of the r-pro-
nouncing prestige pattern. The vertical axis represents average index scores for
 Phonological Correlates of Social Stratification 171
(r): these scores are the percentages of all words with historical (or ortho-
graphic) r in final and preconsonantalposition, in which the speakers used some
form of constricted, consonantal /r/.8 The horizontal axis shows the series of
contextual styles, ranging from the most informal at the left to the most formal
at the right. At the extreme right we have one additional style, D': this is the
subcategory of isolated word lists in which the full attention of the informant
is directed to (r), in such minimal pairs as dock vs. dark, god vs. guard, sauce
vs. source. (The use of (r) in such pairs as nearer vs. mirror is not included in
D', but is includedin D.)
In Figure 2, the population has been divided into six distinct class groups,
and the values of the index scores for (r) are plotted for each group in each con-
textual style. On the whole, this pattern may be described as fine stratification,
since it appears that the class continuum may be divided into as many small
units as the size of the sample will allow, and correlated accordingly with the
use of (r).
At the extreme left, Style A, only the highest ranking status group shows a

100 Socio-economic
classgroups
SEC lowerclass
1}
4-52 working class
6-8 lowermiddleclass
80 9 uppermiddleclass

0 SEC
60

S40
4-5

20

A B C D D'
casual careful reading word minimal
speech speech style lists pairs
CONTEXTUALSTYLE
FIG.2. Classstratificationdiagramfor (r).
 172 The Ethnography of Communication
significant amount of [Link] other words, in everyday speech, (r)
functions as a prestige marker of the upper middle class alone. However, in suc-
cessively more formal styles, the (r) indexes for the other groups rise more
steeply than that for class 9. In particular, the behavior of the lower middle
class, group 6-8, shows an extremely sharp increase in (r). Whereas this group
is essentially r-less in everyday speech, it shows an average (r) index for the
two most formal styles which is considerablygreater than that of the upper mid-
dle class, reaching almost 80 per cent of maximum. This crossover pattern ap-
pears at first sight as a deviation from the regular correlation of linguistic and
social [Link], a similar crossoverappears in several other class strati-
fication diagrams-in fact, for all those which are involved in processes of lin-
guistic change.
Figure 3, for example, is the class stratification diagram for the variable
(eh). This variable concerns the height of the vowel in bad, dance, ask, half,
etc.-words which in other dialects contain a short low front vowel /za/. The
variable (eh) is identified in that subcategory of words containing the dia-
phoneme /a/ which are monosyllabic morphemesending in voiced stops, voice-
less fricatives, and nasals /m/ and /n/, and the derivatives of these words. The
index is constructed from the following scale of height:

phonetically level with the vowel of

(eh-1) [I:;, 1::, e':'] beer
(eh-2) [e:a, e:a] bear
(eh-3) [a•e~]
(eh-4) [e:] bat
(eh-5) [a :] EasternNew Englandask
The index score is computed by taking the average value of all occurrences of
(eh), and multiplying by 10. In Figure 3, the vertical axis represents average
(eh) scores, and the horizontal axis, the series of contextual styles. Here again,
we see a regular correlation of linguistic behavior with class stratification, with
the lower middle class group 6-8 crossing over the upper middle class in the most
formal style.
In Figure 3, it may be seen that the upper middle class reverses the direction
of phonological shift in most formal style, moving back in the direction of (eh)-
30 for the isolated word list bat, bad, back, bag, batch, badge, bath, etc. This
retrograde movement is accounted for by the presence of a certain number of
upper-middle-classinformants who are able to maintain the intermediate allo-
phone of (eh-3), the raised [ael]. This is the characteristic form of the older
prestige pronunciation, and also of the college-educatedr-pronouncingspeakers
who enter the city from other northern regions. However, for the majority of
speakers who were raised in New York City, there is no intermediate allophone
between (eh-2) [e:a], level with where and bear, and (eh-4), [a: ], a lengthened
and tense form of the vowel in bat. Whereas many upper-middle-classspeakers
[but not all] aim at (eh-3), all lower-middle-classspeakers are governed by the
norm of (eh-4); as Figuie 3 shows, this group comes very close to consistent
 Phonological Correlates of Social Stratification 173
(eh-4) pronunciation. Oscillations which do occur in Style D for lower-middle-
class speakers are always (eh-1) or (eh-2), that is, [bet, bae:d,baek,bae:g,
baeti,
be: adz, bee:0. . .]
The shibboleth of (eh) is even more prominent in the pronunciation of
working-class speakers, though they show less consistency in achieving the
norm of (eh-4). Indeed, it may be said that the pronunciationof (eh) is the one
phonologicalvariable that is uppermost for those working-classspeakerswho are
striving for careful, correct speech.

OTHER LINGUISTIC VARIABLES

The linguistic structures which have been illustrated by the three variables
discussed are similar to those seen in the tabulations of other variables, such as
(dh) and (oh). In the case of (oh), we have a curvilinear pattern, since the
lower class and the upper middle class share the same average values, while the
two central class groups show the most extreme values in casual speech. The
lower class does not participate in either social or stylistic patterns of variation

15
Socio-economic class groups
SEC 0-2 lower class
3-5 working class
6-8 lower middle class
9 upper middle class
20

25
CSEC
0-2

30 3-

6-8

35

40
A B C D
casual careful reading word
speech speech style lists
CONTEXTUAL STYLE
FIG. 3. Class stratification diagram for (eh).
174 The Ethnography of Communication
for this variable; (oh) has not, as yet, attained any social significance for this
group. There are also phonological variables which show only social variation,
but not stylistic variation for the population as a whole: that is, they have not
yet become the objects of overt social affect. The vowels of my and mouth fall
into this category. These interpersonal variables have near-constant values in
the speech of given individuals, but group averages show sharp social stratifica-
tion in response to a larger, underlying process which continues to differentiate
the speech of the different strata.
The data given above is limited to the measurement of objective usage of
the phonological variables. It is also necessary to explore the subjective evalua-
tion of these features, and to isolate the subjective, unconscious reactions to
individual phonological variables. In other studies, derived from the survey of
the Lower East Side, it has been possible to determine whether these variables
actually serve as indexes that enable the average New Yorker to identify the
class position of other New Yorkers, within a reasonablerange of error. More im-
portantly, it appearsthat the great diversity in the use of the variablesis matched
by an extraordinaryunanimity in subjective evaluation which defines New York
City as a single speech community.

LINGUISTIC CORRELATESOF ETHNIC STRATIFICATION

In New York City, racial and ethnic groups play an important part in the
dynamics of social change, and one would expect that their influence would be
reflected in linguistic behavior. The opposition of Jewish and Italian groups
among our informants is reflected in a long series of parallel and opposing treat-
ments of the variables (eh) and (oh). However, there is some reason to think
that this older contrast is giving way to an opposition which follows racial lines
more closely, and in which Negro and Puerto Rican groups are opposed to the
rest of the community. Associated with this differentiation is a very different
type of linguistic index, which is worth consideringbriefly here.
There are certain marginal phonemic differencesin the speech of New York-
ers which serve as near-absolutedifferentiatorsof racial groups. One is the pres-
ence or absence of a phonemic contrast of /i/ and /e/ before nasals. For the
great majority of Negro speakers in the city, whether they come from three gen-
erations of New Yorkers or directly from the South, there is no contrast between
pin and pen, tin and ten, since and cents. This phonemic feature is a regular
characteristicof Southern English, and is common in Border States beyond the
usual limits of southern linguistic features. However, there are no white speak-
ers in our sample raised outside the South who show any tendency towards this
phonemic merger. On the other hand, the merger has reached the stage of abso-
lute regularity among younger Negro speakers. The few exceptions which do oc-
cur are always connected with some prominent circumstance in the life history
of the individual that plainly accounts for the deviation. In two cases, the ex-
ceptions were Negro speakerswho had grown up in Staten Island or in the Bronx
without any Negro friends at all. In two other cases, the informant told of a
 Phonological Correlates of Social Stratification 175
grade school teacher who had stubbornly drilled the entire class in this particu-
lar distinction.
A similar differentiation of Puerto Rican speakers can be found in the ab-
sence of the phonemic differentiation between "tin can" and "I can" as
[ke:an] and [kaen]. Those Puerto Rican speakerswho have the most native New
York City pronunciation will still fail to differentiate the auxiliary from the
noun. Other New Yorkers do make this differentiation;in formal contexts a cer-
tain number have corrected their native [e:a] to [e: ] but the phonemic distinc-
tion between the auxiliary and the noun is never reversed,and eventually emerges
in natural speech in all but a tiny minority of cases.

CONCLUSION
There is an impressive regularity in the forces which are reflected in the vari-
ous phonological correlates used here, forces which move the speech of large
classes of people in such uniform directions and magnitudes. But it is the great
utility of the correlates which should be stressed. The data which are illustrated
here represent only the first step in establishing objective distribution of lin-
guistic features and delineating class norms. This basis may then be utilized to
measure the degree of oscillation for individuals and class groups, measures
which in turn can be correlatedwith social mobility and social insecurity. Sub-
jective reactions, both unconsciousand deliberate, are a part of the overall struc-
ture of linguistic behavior which rests upon these objective [Link] struc-
ture of New York City English is eventually to be seen as a many-dimensional
complex, in which social and stylistic variation intersect with purely linguistic
models of covariation.
With the firm empirical base provided by quantitative measures of linguis-
tic performance,we are in a position to observe linguistic change as it is taking
place, contrasting one generation, or even half-generation,with another. We can
trace such changes along many dimensions of the linguistic structure of the
speech community, and thus approach some of the most central problems of the
mechanismof linguistic evolution.

NOTES
'1 am particularlyindebtedto Dr. Lloyd Ohlinof the New York School of Social Work,
and Dr. Wyatt Jones, of Mobilizationfor Youth, for permissionto use the 1961 survey as
a base,and for theirhelp and cooperationat manystagesof this work.
2 Stable arrays of social and stylistic variation can be derived from subgroupsof the
sample containing as few as 25 respondents,since between five and ten respondentsseem
to give a reproducibleresult for a given class group. However, the largersampleof 81 New
York City respondentsto be discussedhere allows the investigationof superimposedpatterns
on the basic patterns (such as the crossoverpattern discussedbelow), and of other variables
such as racialand ethnic group membership,age level, and sex. On the whole, it may be noted
that linguisticbehavior seems to show much greaterregularity,and requiresmallersamples,
than manyothertypes of socialbehavior.
'The originalMFY index utilized the educationof the breadwinnerfor the class status
of the respondent;in the final analysis,both methodsreachedthe sameresult.
176 The Ethnography of Communication
'An index which combinedtwo indicators-occupation and education--was usefulin
analyzingthose linguisticvariableswhichwere apparentlyestablished at a relativelyearly
periodin the respondents' [Link] of the completereport(Labov1964),is
devotedto the analysisof the socio-economic index,and the examination of individual
indicators.
' The primaryfocus of attentionis the respondent's residencefromthe age of five to
the age of 13. It is apparently in this pre-adolescentperiodthat dialectcharacteristics are
mostfirmlyfixed,andtheautomatic, motor-controlledresponses established.
"
Immunityfromconsciousdistortionis not required, but total suppression removesthe
variablefromthe rangeof measurable The use of ain't,for example,is suppressed
quantities.
so completely in mostcontextualstylesthat it is hardlyusefulfor workof [Link]
otherhand,attemptsto increasethe amountof r-pronunciation have little effectuponthe
valuesof theindexesshownhere.
"Syntacticand morphological indexeshave considerable importancebecauseof their
prominence in socialconsciousness,but theirlow frequencyand susceptibility to conscious
suppression makethemmuchlessusefulfor quantitative [Link] cuesarefrequent
and not easilyalteredin formalsituations,but at presentwe lackthe largebodyof theory
and practicein codifyingintonationwhichwe havefor [Link] the sta-
tisticalmeasures usedin psycho-linguisticresearch,suchas the type-tokenratio,maybe quite
revealing as appliedto this data,but the volumeof calculations requiredmakesthemfar less
efficientfor the analysisof linguisticbehaviorthanthe indexespresented [Link] is doubtful
thatwe canlearnas muchaboutsocialprocesses fromthe studyof ad hocstatistical measures
as we can fromindexeswhichhavesocialmeaningin themselves.
8 Onegroupof wordsis excluded, andutilizedfor a separateindex:thosein which/r/
followsa stressed,midcentralvowel, as in bird,work,[Link] rapidextinctionof the
traditionalvowel /ay/ in thesewordshas led to an earlyand relativelyconsistentuse of
/r/ in thesewordsby mostsectionsof thepopulation.

REFERENCES CITED

BABBITT, E. H.
1896 The English of the lower classesin New York City and vicinity. Dialect Notes
1: 457-64.
YAKIRA
FRANK, A.
1948 The speechof New [Link] Michigan
dissertation.
AnnArbor,
Michigan.
ALLAN
HUBBELL, F.
1950 The pronunciation
of Englishin New York City. New York,King'sCrown
Press,ColumbiaUniversity.
HYMES, DELL
1962 The ethnographyof [Link] Anthropologyand Human Behavior,Thomas
GladwinandWilliamC. Sturtevant,[Link],
D. C., TheAnthropologi-
cal Societyof Washington,
pp. 15-53.
KURATH, HANS and RAVEN A. McDAVID, JR.
1961 The pronunciationof English in the Atlantic States. University of Michigan
Press, Ann Arbor, Michigan.
LABOV,W.
1963 The social motivation of a sound change. Word 19: 273-309.
1964 The social stratification of English in New York City. New York, Columbia
University dissertation.
MICHAEL, JOHN W.
1962 The construction of the social class index. In Appendix A-i, Codebook, Mobili-
zation for Youth, 214 East 2nd St., New York, N .Y.