Sociolinguistics — variation, identity, and language in society

Intuition Beginner

Sociolinguistics is the study of how language varies across speakers and settings, and of what that variation does. A New Yorker, a Texan, and a Londoner each say "car" differently. Those differences do not change the meaning of the word. They carry something else: where the speaker is from, and how they position themselves socially.

The single guiding idea is that how you speak signals who you are. Drop a consonant, raise a vowel, switch a word, and you mark region, class, age, gender, ethnicity, or how formal the moment feels. Speakers choose among variants — most of the time without thinking — and listeners read those choices as social information.

Sociolinguistics asks four questions about this variation. Who uses which variant? When and why do they switch? What do listeners infer from a variant? How do these patterns change over generations? Phonology 51.02.01 gave the sound categories. This unit asks why real speakers do not use them uniformly.

Visual Beginner

A stratification chart: three New York department stores ordered by social class, with a bar for the percentage of speakers pronouncing the final "r" in words like "fourth floor."

The higher the store's social ranking, the more often shoppers and staff pronounce the final "r." The same sound variable tracks social position.

Store (1962 NYC)	Social level	Final /r/ pronounced
Saks Fifth Avenue	high	62%
Macy's	middle	51%
S. Klein	low	20%

This is William Labov's classic department-store result. The variable (whether you say the "r" in fourth floor) is not random: it rises with social class.

Worked example Beginner

Use Labov's 1966 New York City study. The variable is final /r/ — whether speakers pronounce the "r" in fourth floor, or drop it to sound like fawth flaw.

Step 1. Pick three stores that differ in social ranking: Saks (high), Macy's (middle), S. Klein (low). Ask each shopper a question whose answer contains /r/ words, and note whether they pronounce the consonant.

Step 2. Count tokens. At Saks, about 62 of every 100 possible /r/ sounds were pronounced. At Macy's, about 51. At S. Klein, about 20.

Step 3. Read the pattern. The percentage climbs with the store's social level: 20 to 51 to 62. The sound variable is stratified by class.

Step 4. Labov then asked speakers to repeat the phrase "carefully." The /r/ rate rose in all three stores. So speakers can shift style upward when they pay attention to speech.

What this tells us: a single sound variable encodes both social class (between stores) and formality (within a speaker). Variation is structured, not random.

Check your understanding Beginner

Formal definition Intermediate+

A sociolinguistic variable is a linguistic item with two or more variants whose relative frequencies pattern with non-linguistic factors — speaker group, style, setting, or time ^{[Labov 1972]}. A variant is a concrete way of realising the variable: pronounced /r/ versus dropped /r/; singing versus singin'; y'all versus you guys. The empirical object is the distribution of variants across contexts, not any single occurrence.

The index score of a binary variable for a speaker group $G$ is

$$I_G=100\cdot \frac{n_G^{+}}{n_G^{+}+n_G^{-}},$$

where $n_G^{+}$ is the token count of one variant and $n_G^{-}$ the count of the other ^{[Labov 1966]}. The index is one hundred times the group's relative frequency of the chosen variant, so it lies in $[0,100]$. For Labov's (r), taking $+\mathrm{v}$ to mean "pronounced," Saks scores $I\approx 62$, Macy's $\approx 51$, S. Klein $\approx 20$.

A speech community is a group of speakers who share not an identical grammar but shared norms of evaluation about language — they agree, for instance, that one variant sounds "more correct" or "more friendly," even when they use different variants themselves ^{[Gumperz 1968]}. A dialect is a variety associated with a region (regional dialect) or a social group (social dialect, or sociolect). Register names variation by setting or topic (legal, academic, casual); style names variation by degree of attention and formality within a single speaker's repertoire ^{[Wardhaugh & Fuller]}.

Prestige is the social value attached to a variant. Overt prestige attaches to the standard or high-status form, the one taught in school and used in formal settings. Covert prestige attaches, within a subgroup, to a non-standard form that signals local loyalty or toughness ^{[Trudgill 1974]}. The two can pull in opposite directions: a speaker may use a standard variant at work and a vernacular one at home, each carrying prestige in its setting.

Diagnostics: indicator, marker, stereotype

Labov sorted variables by how much social awareness they carry.

• Indicator. Shows stable social differentiation but no style-shifting: speakers use it the same way in casual and careful speech. Listeners rarely comment on it.
• Marker. Shows both social differentiation and style-shifting: the variant rate moves with attention to speech, as /r/ did. Listeners are dimly aware of it.
• Stereotype. Is the object of overt social commentary and often of caricature ("Brooklynese" toity-toid for thirty-third). Stereotypes are typically already receding in real use.

The three form a gradient of consciousness, from invisible differentiation through style-sensitive marking to overt naming.

Linguistic theory Intermediate+

Variationism and the sociolinguistic interview (Labov). The variationist method treats language as fundamentally heterogeneous. Where earlier dialectology collected one "best" speaker per village, Labov sampled many speakers, coded many tokens, and read the distribution of variants as the signal. The sociolinguistic interview is the controlled setting: a chain of topics from casual (childhood games, danger of death) to careful (word lists, minimal pairs) that lets the interviewer read off a speaker's style range ^{[Labov 1972]}. The central claim is the principle of structured heterogeneity: variation is not noise but a system, with regular constraints from class, style, age, sex, and ethnicity.

Communicative competence (Hymes). Against the Chomskyan picture of a homogeneous ideal speaker-hearer with "linguistic competence," Dell Hymes argued that what speakers know is communicative competence: when to speak and when to stay silent, what to say to whom, and how to interpret the social force of an utterance ^{[Hymes 1972]}. The unit of analysis is the speech event and its ways of speaking, not the isolated sentence.

Two contested questions

Contested question 1 — Is the speech community homogeneous or heterogeneous?

Position 1 — Homogeneity of norms (Bloomfield 1933; Gumperz 1968). A speech community is a group united by a common language and by shared norms of interaction and evaluation. What holds speakers together is agreement about what counts as the same, what counts as polite, what counts as correct; the community is defined by these shared norms, and uniformity of production is not required. On this view the community is a real, cohesive social object, and the analyst's job is to recover its norms ^{[Gumperz 1968]}.

Position 2 — Heterogeneity and the community of practice (Labov 1972; Eckert 2000). Communities are cross-cut by class, gender, age, ethnicity, and local affiliation, and speakers affiliate not with an abstract "community" but with a community of practice: a group brought together by mutual engagement in a shared enterprise (a high school, a factory shift, a gospel choir). Variation is the means by which speakers construct and negotiate these local identities. Eckert's Belten High study showed that the same vowel variable indexed different things for "jocks" and "burnouts," and that identity, not demographic category, ordered the variation ^{[Eckert 2000]}.

How to read the dispute. The two positions name different grains of analysis. Gumperz's community describes the normative level — the standards speakers orient to and judge by. Eckert's community of practice describes the produced level — the variants speakers actually deploy to build identities. Most working sociolinguists use both: shared norms explain why listeners agree that one variant "sounds educated," while communities of practice explain why two boys in the same school sound different.

Contested question 2 — Does language shape thought? (The Sapir-Whorf debate.)

Position 1 — Strong linguistic determinism (Whorf 1956). The grammar and lexicon of a language determine what its speakers can think, so that speakers of languages with different categories inhabit different cognitive worlds. On the strongest reading, a language without tense markers leaves its speakers unable to reason about time in the way speakers of tensed languages do.

Position 2 — Weak linguistic relativity and the universalist reply (Lucy 1992; Pinker). Weak relativity holds that language influences habitual thought — directing attention, easing recall, biasing categorisation — without fixing the bounds of the thinkable. Empirical work (colour categorisation, spatial frames of reference) finds small, context-bound effects rather than deep cognitive imprisonment. The universalist reply, voiced by Pinker and others, holds that the core conceptual repertoire is shared across humans and that language is one channel among many for thought, not its mold.

How to read the dispute. The strong deterministic claim has fared badly under experiment; the weak, probabilistic claim is the live one, and it is exactly the kind of claim sociolinguistics and linguistic anthropology 31.05.01 test with matched populations and controlled tasks.

Bridge. This result builds toward 51.06.01 (historical linguistics), where the actuation problem asks why a sound change ignites in one community and one decade rather than another, and appears again in 30.04.01 (social stratification), whose class, status, and ethnicity variables are the independent variables of variationist analysis; the foundational reason variation is structured is that speakers deploy variants to index social meaning, the central insight is that language is heterogeneous in a principled way, and this is exactly the architecture whose consequences the historical and stratification units then unfold.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none. Sociolinguistics is an empirical science of social meaning adjudicated by corpus counts and social survey, not by formal proof. Its core notions — the sociolinguistic variable, the speech community, overt and covert prestige, the community of practice, and language attitudes — have no primitive in Mathlib, and its central disputes (homogeneity versus heterogeneity of the speech community; the strength of linguistic relativity) are empirical and conceptual questions rather than theorems. The one nearby formal object, the Bernoulli-model index score $I=100n^{+}/(n^{+}+n^{-})$, reduces to elementary maximum-likelihood probability already present in Mathlib and is worked out in the Master section below; it is a statement about a statistical estimator, not about the linguistic claims of this unit.

Advanced results Master

The three social embeddings of a variable (Labov). Labov showed that a variable can pattern with social class (stratification), with style (attention), and with change over time (age-grading or a change in progress). Plotting a variable on all three axes at once is the heart of the variationist display. When the lines for social classes are parallel and separated, the variable is a stable marker of stratification; when the lines cross or fan out across age cohorts, a change is in progress, and the youngest cohort's curve points toward the future state of the language ^{[Labov 1966]}.

The curvilinear principle. For a change in progress from below the level of conscious awareness, the interior social groups — the upper-working and lower-middle classes — lead the change, with both the highest and lowest groups lagging. The frequency curve bends like an arch over the class axis. This is diagnostic: a monotone rise or fall across class signals a stable stratification (as with /r/), whereas a curvilinear peak in the middle signals an active change (as with the raising of (aw) in Labov's later NYC work). The principle constrains how innovations propagate: the social extremes are not the usual innovators ^{[Labov 1972]}.

Sex differentiation and prestige (Trudgill 1974). In Norwich, Trudgill found that women used the standard variant more than men of the same class for nearly every variable, while men over-reported their use of vernacular variants. Trudgill traced this to the dual prestige system: women orient to the overt prestige of the standard (a proxy for status in a society where women's status is less secure), while men are drawn to the covert prestige of the working-class vernacular, which signals toughness and local solidarity. The sex difference is thus a difference in which kind of prestige a group orients to, not a difference in linguistic capacity ^{[Trudgill 1974]}.

Social networks (Milroy 1980). Lesley Milroy measured speakers' integration into local networks with a network-strength scale counting ties of kinship, work, and neighbourhood. Speakers high on the scale preserved Belfast vernacular vowels; speakers low on the scale converged toward the standard. Dense, multiplex networks — where the people you work with are also your kin and neighbours — enforce local norms and resist change. Loose, uniplex networks let outside norms in. Network density thus predicts the rate and direction of change ^{[Milroy 1980]}.

Communities of practice (Eckert 2000). Eckert reframed demographic categories as emergent identities. At Belten High, the vowel variables (ay) and (aw) did not line up with raw class or sex; they lined up with whether a student was a "jock" or a "burnout," and within those, with how central the student was to the group. Variation, on this view, is a semiotic resource: speakers combine variants to stake out positions in a local social landscape. The community of practice, not the census category, is the unit at which meaning is made ^{[Eckert 2000]}.

Language attitudes and the matched-guise technique (Lambert 1960). Listeners hear the same bilingual speaker producing the same text in two languages (or two accents) and rate each "guise" on traits like intelligence, kindness, or competence. Because only the language variety changes, any difference in ratings isolates the social evaluation attached to the variety itself. The method shows that attitudes attach to varieties, not to speakers: the identical person is judged differently in each guise ^{[Wardhaugh & Fuller]}.

Language contact, borrowing, and multilingualism. When speech communities are in sustained contact, speakers borrow words, sounds, and structures, and often code-switch — alternate varieties within a conversation. Borrowing imports lexical items (English café from French); convergence shifts the grammar of one variety toward another under bilingual pressure; pidginisation and creolisation build new contact varieties when speakers with no common language must communicate. Diglossia (Ferguson 1959) names a stable situation in which two varieties divide a community's functions: a High variety for formal writing and speech, a Low variety for intimacy and casual talk. These outcomes are the synchronic surface of the diachronic contact studied in 51.06.01.

Language planning and standardisation. Communities do not merely have a standard; they make one, through four linked activities (Haugen 1966): selection of a base variety, codification in dictionaries and grammars, elaboration of the standard for new domains (science, administration), and acceptance by institutions. Standardisation is a political process: the chosen standard usually reflects the variety of a dominant group, and its prestige is then taught as if it were intrinsically superior. The same process underwrites status, corpus, and acquisition planning in multilingual states ^{[Wardhaugh & Fuller]}.

Synthesis. Putting these together, sociolinguistics reframes language as a social object: the variationist method established here builds toward 51.06.01 where sound change is the diachronic face of structured variation, appears again in 31.05.01 where ethnography asks how speech indexes culture and identity, the foundational reason variation carries social meaning is that no two speakers share an identical repertoire, this is exactly why the speech community is defined by shared norms of evaluation rather than uniform production, and the pattern generalises from phonetics through syntax to the discursive and stylistic levels of language; the bridge is that structured heterogeneity is the common substrate of language change, language contact, and identity work.

Full proof set Master

Proposition (Index score, stratification monotonicity, and the curvilinear test). Let $G_1,\dots ,G_k$ be $k$ speaker groups and let a binary sociolinguistic variable have two realisations, the prestige variant $[+\mathrm{v}]$ and the vernacular variant $[-\mathrm{v}]$. For group $G_i$ let $n_i^{+}$ and $n_i^{-}$ be the token counts of the two realisations, with $N_i=n_i^{+}+n_i^{-}>0$, and define the index score $I_i=100n_i^{+}/N_i$.

(i) $I_i/100$ is the maximum-likelihood estimate of the probability that a randomly drawn token from $G_i$ is $[+\mathrm{v}]$.

(ii) The relation $G_i⪯G_j⟺I_i\le I_j$ is a total pre-order on the groups by frequency of the prestige variant, and it is antisymmetric (hence a total order on equivalence classes) precisely when the $I_i$ are pairwise distinct.

(iii) A variable is stratified across the groups whenever the pre-order $⪯$ has more than one equivalence class; applied to Labov's (1966) department-store (r) data the ordering $I_{\mathrm{Saks}}\approx 62>I_{\mathrm{Mac}{\mathrm{y}}^{\prime }\mathrm{s}}\approx 51>I_{\mathrm{Klein}}\approx 20$ witnesses stratification, whereas a curvilinear pattern $I_{\mathrm{middle}}>I_{\mathrm{high}},I_{\mathrm{low}}$ in the same index witnesses a change in progress rather than stable stratification.

Proof. For (i), model each token as an independent Bernoulli trial with success $+\mathrm{v}$ and failure $-\mathrm{v}$, and unknown success probability $p_i\in [0,1]$. The likelihood of observing $n_i^{+}$ successes in $N_i$ trials is

$$L(p_i)=\left(\genfrac{}{}{0ex}{}{N_i}{n_i^{+}}\right)p_i^{n_i^{+}}(1-p_i{)}^{N_i-n_i^{+}}.$$

Taking logarithms, $\log L(p_i)=n_i^{+}\log p_i+(N_i-n_i^{+})\log (1-p_i)+\text{const}$, and differentiating gives ${\partial }_{p_i}\log L=n_i^{+}/p_i-(N_i-n_i^{+})/(1-p_i)$. Setting this to zero yields ${\hat{p}}_i=n_i^{+}/N_i$, and the second derivative ${\partial }_{p_i}^2\log L=-n_i^{+}/p_i^2-(N_i-n_i^{+})/(1-p_i{)}^2<0$ on $(0,1)$ confirms it is a maximum. Hence $I_i=100{\hat{p}}_i$, proving (i).

For (ii), the map $G_i\mapsto I_i\in [0,100]$ induces the relation $G_i⪯G_j⟺I_i\le I_j$. Totality and transitivity follow from those of $\le$ on the reals; reflexivity is immediate. Two groups are equivalent, $G_i\sim G_j$, exactly when $I_i=I_j$, i.e. when ${\hat{p}}_i={\hat{p}}_j$ — when they use the prestige variant at the same estimated rate. The pre-order collapses to a total order on the equivalence classes precisely when all $I_i$ are distinct, i.e. when no two groups share a rate.

For (iii), stratification across the groups means the index is not constant on $\{G_1,\dots ,G_k\}$, which by (ii) is the statement that $⪯$ has more than one equivalence class. For Labov's (r) data the three values $62,51,20$ are pairwise distinct, so the stores fall into three singleton classes ordered $I_{\mathrm{Saks}}>I_{\mathrm{Mac}{\mathrm{y}}^{\prime }\mathrm{s}}>I_{\mathrm{Klein}}$, matching the social ranking by price level and customer class: the variable is stratified, with the prestige variant (pronounced /r/) more frequent at higher social levels. By contrast, when a single interior group leads — $I_{\mathrm{middle}}>I_{\mathrm{high}}$ and $I_{\mathrm{middle}}>I_{\mathrm{low}}$ — the ordering is not monotone in class rank, so the curvilinear principle flags a change in progress rather than a stable stratification; the shape of the index curve, not its mere separation, distinguishes the two cases. ∎

The proposition formalises the basic move of urban dialectology: reduce a cloud of tokens to a group-level rate, compare rates as a pre-order, and read off both the fact of stratification (distinct classes) and its character (monotone versus curvilinear) from the ordering alone. Everything in Labov's display — class, style, and age curves plotted together — rests on this index-and-order construction.

Connections Master

• Phonology 51.02.01. Every sociolinguistic variable is built on a phonological contrast — a phoneme, a feature, or an allophone — that phonology defines; the variationist counts which realisation of a phonological category a speaker produces, so sociolinguistics presupposes the segmental and featural inventory established there.

• Historical and comparative linguistics 51.06.01. Sound change is the diachronic face of structured variation: a change in progress is a synchronic variable whose distribution is shifting across age cohorts, so the actuation problem (why a change starts where and when it does) is jointly a historical and a sociolinguistic question.

• Social stratification 30.04.01. Class, status, ethnicity, and gender are the independent variables of variationist analysis; the sociological categories that order inequality in a society are the same categories along which language varies, so the stratification axis of a sociolinguistic variable is read directly off the social structure.

• Linguistic anthropology 31.05.01. Ethnographic method, the speech event, and the question of how language indexes culture and identity are shared terrain; communities of practice, ways of speaking, and the relativity debate are investigated jointly from the anthropological and the sociolinguistic side.

Historical & philosophical context Master

The recognition that language varies with its speakers is as old as the study of language. The Sanskrit grammarian Pāṇini registered social and regional variants (chandasi, "in the Vedic language"; desi, "regional") alongside the canonical form, and the Arabic grammarians catalogued Bedouin tribal variants as evidence for correctness. Dialect geography proper began in the late nineteenth century with Georg Wenker's and Jules Gilliéron's atlases, which mapped one form per locality and treated variation as a regional fact to be plotted ^{[Labov 1972]}.

The modern, social turn is William Labov's. In The Social Stratification of English in New York City (1966) and the collected papers of Sociolinguistic Patterns (1972), Labov showed that variation is systematic, that it correlates with class, style, age, and sex in regular ways, and that the speech community is held together by shared norms of evaluation rather than by uniform production. The department-store and Martha's Vineyard studies founded the variationist paradigm and supplied its method, the sociolinguistic interview ^{[Labov 1972]}. Peter Trudgill's Norwich study (1974) fixed the sex-and-prestige pattern; Lesley Milroy's Belfast network study (1980) supplied the network-strength scale; and Penelope Eckert's Linguistic Variation as Social Practice (2000) reframed the field around the community of practice.

Two philosophical fault lines run through the discipline. The first is the speech-community question — whether the community is a normatively cohesive whole (Gumperz) or a contested field of identity work (Eckert) — which decides whether the analyst studies shared norms or local practice. The second is the Sapir-Whorf question of linguistic relativity, which sets the strong deterministic claim of early Whorf against the weak, probabilistic relativity that modern experiment supports. Both remain live, and both are taught here as open disputes rather than settled doctrine.

Bibliography Master

@book{Labov1966,
  author = {Labov, William},
  title = {The Social Stratification of English in New York City},
  publisher = {Center for Applied Linguistics},
  year = {1966},
}

@book{Labov1972,
  author = {Labov, William},
  title = {Sociolinguistic Patterns},
  publisher = {University of Pennsylvania Press},
  year = {1972},
}

@book{Trudgill1974,
  author = {Trudgill, Peter},
  title = {The Social Differentiation of English in Norwich},
  publisher = {Cambridge University Press},
  year = {1974},
}

@book{Milroy1980,
  author = {Milroy, Lesley},
  title = {Language and Social Networks},
  publisher = {Blackwell},
  year = {1980},
}

@book{Eckert2000,
  author = {Eckert, Penelope},
  title = {Linguistic Variation as Social Practice: The Linguistic Construction of Identity in Belten High},
  publisher = {Blackwell},
  year = {2000},
}

@book{WardhaughFuller2014,
  author = {Wardhaugh, Ronald and Fuller, Janet M.},
  title = {An Introduction to Sociolinguistics},
  edition = {8},
  publisher = {Wiley-Blackwell},
  year = {2021},
}

@book{Meyerhoff2011,
  author = {Meyerhoff, Miriam},
  title = {Introducing Sociolinguistics},
  publisher = {Routledge},
  year = {2011},
}

@incollection{Hymes1972,
  author = {Hymes, Dell},
  title = {On Communicative Competence},
  booktitle = {Sociolinguistics},
  editor = {Pride, J. B. and Holmes, Janet},
  publisher = {Penguin},
  year = {1972},
}

@incollection{Gumperz1968,
  author = {Gumperz, John J.},
  title = {The Speech Community},
  booktitle = {International Encyclopedia of the Social Sciences},
  publisher = {Macmillan},
  year = {1968},
}