Variacion y Cambio Linguistico PDF
Variacion y Cambio Linguistico PDF
Languages are not permanent entities. If we have a look at an extract of a text from different periods of the English
language we would see that languages can change a lot over a period of time.
Languages change due to sociolinguistic factors and internal psychological factors that reside in the structure of the
language and the minds of the speakers. Language change occurs due to variation within a language and/or because
of contacts with other languages.
It is not possible to predict language change. However, for centuries scholars have sought explanations for why
certain changes might have taken place or why certain elements might be retained:
1. System internal explanations > language is a system in which internal forces operate and lead to change
2. Psychological explanations
3. Geographical explanations
4. Sociolinguistic explanations > language changes because the speakers are social beings and use different
linguistic variants depending on their role in society.
Some research topics related to language change include local dialects, standard languages, sociolects, difference
between spoken and written language, among others.
For language change to take place there must be two equivalent linguistic forms, one giving way to the other
eventually. Sociolinguists distinguish between:
• Changes from above > changes we are consciously aware of and also to linguistic items, mainly new sounds
found in prestigious varieties and introduced by the dominant social class.
• Changes from below > changes we are not fully aware of, and also less prestigious varieties influencing
more prestigious ones.
Language change can occur on every linguistic level. So we have phonological, morphological, syntactic, lexical and
pragmatic changes. Some examples of change are: new words can appear through loan, a word can acquire new
meanings, among others. Linguistic elements can also change their status within the linguistic system. The process
of a lexical word changing into a grammatical element is called grammaticalization.
The study of language change has altered its character in recent years. Scholars have shifted from reconstructing the
earliest stages of languages (comparative method) and paying no attention to changes currently taking place, to
using Sociolinguistic methodology to study changes that took place in the past.
Linguistics as we know it today appeared at the beginning of the 20th century and developed independently in Europe
and America.
On the one hand, Europeans had a philosophical tradition and a knowledge of the historical study of languages based
on the comparative philology. Studies focused on the evolution of classical languages and to a lesser extent on
modern European languages through the analysis of written testimony.
In America, linguistic research focused on Amerindian languages for which no written records were kept, so that it
was impossible to develop an approach based purely on written texts. It was decided, therefore, to carry out an
anthropological study of languages.
In the 19th century, there is a new perception into the study of Linguistics influenced by Darwinian studies in which
the language is considered as an organized system. Besides, history and historical comparison are introduced in
Linguistics. Comparative studies between languages sought regularities in order to reconstruct the Proto-Indo-
European language.
The second half of the 19th century witnessed a major success of historical-comparative linguistics. At this time the
Neogrammarian School was founded. It developed the theory that language change is systematic and regular and is
governed by a set of rules. It is based on two fundamental principles:
1. Phonological change takes place according to mechanical laws that have no exceptions (principle of
regularity)
2. Importance of analogy in the creation of new forms.
However, limitations in Comparative and Historical Linguistics were soon discovered, since not all exceptions can be
explained. For example, the method assumes that when two languages are separated from the mother tongue, they
stop influencing each other, which is not entirely true. Moreover, Comparative Linguistics is interested in spelling
rather than sounds. In this sense, in the 20th century there was a reaction against purely linguistic studies and in
favour of the study of speech variation.
Both the typological and the genealogical classification of languages ignore the importance of the cultural
relationship between languages, the fact that languages influence one another by contact between them.
Sometimes it is impossible to know if two languages are similar because they share a common origin or because of
the contact between them. These limitations led to a new approach to the study of languages, Structuralism, which
studies language as a structure, a system that allows changes.
1957 was the beginning of Generative Linguistics with the publication of Noam Chomsky’s Syntactic Structures.
Linguistics shifts to a psycho-biological stage, with interest in the way in which children acquire languages on the
basis of an abstract universal grammar common to all languages.
2.2 Sociolinguistics
Sociolinguistics’ aim is to study the social uses of language, language as a social phenomenon, and also language
diversity/varieties. It started in the 60s as an opposition to Chomsky’s idea. Sociolinguistics is interested in language
use within human societies since it is believed that there is a close relationship between language and society, and
that language reflects social patterns and divisions. Sociolinguists concentrate more on synchrony.
In Europe, Sociolinguistics began with the study of Historical Linguistics and Linguistic Geography. In America, on the
other hand, the study of Sociolinguistics arises from contact of linguists with other disciplines such as Anthropology
and Sociology. Sociolinguistics today is divided into subfields such as Pragmatics, Studies of Pidgins and Creoles,
Planning Studies and Language Policy, Historical Sociolinguistics, among others.
Sociolinguistics has fostered a new vision of linguistic variation and change as inherent components of languages
and has a highly interdisciplinary nature as it can be applied to various fields of research. Sociolinguistics has also
contributed greatly to the advancement of Historical Linguistics, since the study of linguistic variation can help
understand historical phenomena and vice versa, indicating that through linguistic phenomena studied in the
present we can explain those changes that took place in the past.
Major sociolinguistic studies in recent decades have focused on determining the social evolution of language
variants. The study of the social use of language began in 1963 when Labov presented the first report of
sociolinguistic research, in which he introduced new techniques in sample surveys, natural experimentation and
quantitative analysis in linguistic research. Labov studied for the first time language change in progress.
In 1968 the linguists Weinreich, Labov and Herzog published an article in which the studied language change and
proposed social motivation as its most influential cause. Sociolinguistics leaves a space open to the individual who
uses that instrument in their interactions with others, not only to communicate but also to define themselves.
Language should be understood as a social phenomenon that establishes a relationship between the speakers and
their social environment, and among other members of the linguistic community.
Two undeniable contributions of Sociolinguistics to the study of language change are the careful and accurate
observation of the present variants, methods of data recording, and the analysis and evaluation of the data using
the latest technology. These results can be extrapolated to the past, following the uniformity principle adopted from
modern Geology.
In the 70s and 80s, Peter Trudgill began to study language contact, especially dialects in contact.
In this course we are concentrating on the variationist approach to Sociolinguistics. The basic methods in variationist
studies are the following:
English Historical Sociolinguistics is a discipline that emerged in the 80s inspired by Labov’s work, which consists in
applying Sociolinguistic methodology to historical data, explaining variation and change in the history of English.
Historical Linguistics is concerned with studying how languages change over time and developing and implementing
methods for reconstructing early stages of a language based on written records that have been preserved over the
centuries. Present-day sociolinguistic methodology can be employed to historical data, to explain variation and
change in the history of English.
Historical linguists are both historians and linguists. They are responsible for collecting reliable data and evaluating
the sources. They often have to face the problem of studying periods for which historical records are not
documented.
Historical Linguistics has changed in recent years. Traditionally linguists engaged in reconstructing the earliest stages
of the language and describing phonological changes overtime. They did not pay much attention to changes from a
synchronic point of view, to syntactic changes, to pidgin and creole languages, language death, or sociolinguistic and
psycholinguistic factors underlying many alterations or changes.
In the last thirty years all these issues are receiving a lot of attention. Historical Linguistics has experienced a
methodological and conceptual transformation in recent decades, and it has been enriched by contributions from
other disciplines such as Sociolinguistics, Corpus Linguistics, studies in language change, and Contact Linguistics.
Historical Sociolinguistics appears as a new discipline with similar aims to those of Sociolinguistics. The main
difference with Sociolinguistics lies in that the application of methods of comparative and internal reconstruction
are essential in Historical Linguistics, but not in studies of variation. Synchronic Sociolinguistics often focuses on the
study of variation and change at the phonological level, while Diachronic Sociolinguistic often focuses on the
grammatical level.
Historical Sociolinguistics has advanced dramatically in the last twenty years with a special mention to the Finish
linguists who have been applying the principles of Sociolinguistics to English historical studies since the 90s. In 1991,
the Academy of Finland funded a project whose main aim was to compile a corpus to facilitate sociolinguistic
research applied to the history of English.
This project resulted in the Corpus of Early English Correspondence whose objective was to apply sociolinguistic
methods to historical data. This corpus is composed of letters that were selected systematically between men and
women, young and adult, from different social strata. They also attempted to cover all the regions of England,
although most of the rural population was illiterate. Attention was also paid to the contents of the letters, ranging
from love letters to family or business matters.
Historical Sociolinguistics faces a number of problems. The data available to historical sociolinguists consists mainly
of texts written in the past that have been preserved by chance. No text should be considered as a historical witness
if it contains emendations, modernization, alteration of the scribe’s punctuation, any attempt at reconstructing
through several versions, or any form of standardization of spelling variables. The study of any past stage of language
is problematic, because the only witness is the written text. These texts present variation and occasionally the textual
language is not necessarily equivalent to the spoken language of the scribe. Furthermore, the texts are often
incomplete and isolated from the context in which they originated. Moreover, in many cases not all styles, registers
or varieties are represented in the texts preserved.
Corpus Linguistics emerged in the late 60s. Text corpora offer many possibilities to researchers: we can count a word
frequency in one or various texts, we can search different types of collocations, etc.
English Historical Linguistics, for more than a decade, has been benefiting from its potential. There is now a clear
interest among historical linguists in the use and creation of electronic corpora for research because they see them
as a source of information to study the changes in the language over time. For a long time, the main problem for
researchers was the lack of representative historical corpora that could be used in the diachronic study of language.
However, today the number of computerized text corpora is growing rapidly. This seems to be a natural evolution
as historical linguists have always based their research on collections of texts from early periods to document or
study language change. As a result, many historical linguists have shifted their focus of attention to research based
on electronic corpora.
Historical linguists need corpora containing texts from different genres and periods. Diachronic corpora can be used
to study lexical, grammatical, dialectological and registers features (e.g. how the language of men and women has
changed over the centuries). The computerized corpora are one of the most widely used tools in modern
Dialectology.
The best known English historical corpus is the Helsinki Corpus of English Texts containing about 1.6 million words
dating from the Early Old English period (850) until the end of the Early Modern English period (1710). The corpus
covers a wide range of genres, regional variations and sociolinguistic variables such as sex, age, education and social
class. Although this corpus is not tagged, each text contains information about the author, date of composition and
dialect.
UNIT 1 GLOSSARY
Changes from above – It refers to language change that we are consciously aware of and also to linguistic items,
mainly new sounds found in prestigious varieties and introduced by the dominant class.
Changes from below – It refers to language change that involves changes we are not fully aware of, and also less
prestigious varieties influencing more prestigious ones.
Conscious and unconscious change > There are pressures from above which are conscious changes which means
that individuals know that they are changing their language. There are also pressures from below which are seen
as unconscious changes where individuals subconsciously are changing their language meaning that they are not
aware of their own actions.
Diachrony > It refers to the way in which a language has developed and evolved through time. This evolution is
then seen from a historical point of view.
Dialect > It is a regional or social variety of a language distinguished by pronunciation, grammar, and/or
vocabulary. It is often used to characterize a way of speaking that differs from the standard variety of the
language.
Externally motivated language change > It refers to a change which would appear to be triggered and guided by
social considerations.
Grammaticalization > It is the process of a lexical word changing into a grammatical element.
high prestige variety > It refers to a king of variety of language whose speech habits are generally considered
positive among any linguistic community.
Historical Linguistics > It is the study of language change through a diachronic point of view. It describes and
accounts for observed changes in particular languages, develops theories to explain language changes and
reconstructs the early stages and history of languages and determine their relatedness.
Historical Sociolinguistics > It is a discipline that consists in applying Sociolinguistic methodology to historical data,
explaining variation and change through the history of a language. It focuses on the socio-political context of the
evolution of languages rather than on a descriptive and detailed documentation of the individual forms of a
language.
internal vs external change > Internal changes are those realized by psychological factors that reside in the
structure of the language and the minds of the speakers. External changes are those realized by sociolinguistic
factors (social background, gender, age, context, etc.)
Internally motivated language change > It refers to any change which can be traced to structural considerations in
a language and which is independent of sociolinguistic factors.
Lexical diffusion > It is a phenomenon that occurs when a phonological rule gradually extends its scope of
operation to a larger and larger portion of the lexicon, until all relevant items have been transformed by the
process.
Loanword > It is a word imported into one language from another language with little or no modification.
Low prestige variety > It refers to a kind of variety of language whose speech habits are generally considered
negative among any linguistic community.
Principle of regularity > It is a phonological principle in language change developed by the Neogrammarian School
that claims that phonological change takes place according to mechanical and systematic laws that have no
exceptions.
social network > It relates to a particular speech community in terms of relations between members in that
community. In it, language is embedded.
Sociolinguistic variable > Social factors that affect the linguistic patterns of a given community of speakers and
make them produce different realizations of saying the same thing. Sex, age, education or social class are examples
of it.
speech community > It describes a distinct group of people who use language in a unique and mutually accepted
way among themselves. They can be members of a profession with a specialized jargon, distinct social groups like
high school students, or even tight-knit groups like families and friends. Its members will often develop slang or
jargon to serve the group’s special purposes and priorities.
Standard language > It is a language variety broadly spread and generally understood by the whole community of
speakers of a given language. It is often used in formal education and also widely used by mass media.
Synchrony > It is concerned with the study of language as it exists at one point in time, that is to say, the study of
linguistic phenomena only at a given time, typically the present.
Synchronic Sociolinguistics > It studies language change currently taking place. It is the study of language change
by using quantitative models and different sociolinguistic methods. It often focuses on the study of variation and
change at the phonological level. It also aims at demonstrating how language changes that take place in the
present can explain those changes that once took place in the past, as the synchronic study of linguistic variation
can help understand historical phenomena.
Variation > It refers to regional, social or contextual differences in the ways that a particular language is used.
Variationist approach > It refers to a method that examines the patters of a language that are a variable from
speaker to speaker or from context to context.
Variety > It refers to a specific form of language or linguistic expression.. This may include languages, dialects,
accents, registers, styles or other sociolinguistic variation, as well as a standard variety itself.
QUESTIONS ABOUT LANGUAGE AND VARIATION, AND WHERE WHE GOT THEM
This book is about the study of language and the search for explanations of why languages are spoken the way they
are. In the perspective taken in the book, the goal of variationist linguistics is to understand why languages varieties
become different from one another (or, in some cases, become more similar).
We will take a historical perspective, because the theories, questions and ideas about language discussed in the
variationist field are not the result of an historical process, but, even though it is a ‘new’ field, are owed to the
questions formulated decades and centuries ago.
The roots of the variationist programme may be traced to observations made by Sir William Jones, who is often
credited as the first European to remark, in 1786, that the similarities among Sanskrit, Greek, and Latin suggest that
they somehow derived from a common source. This insight led to the development of the comparative method in
historical linguistics, and the 19th century showed a rapid development in the understanding of relationships among
languages. In addition to this discovery, one of the intellectual currents in European society about this time was a
move towards rationalism and the discovery of ‘natural’ laws based on direct observations. Comparative historical
linguistics, then, aroused out of the same trends in science that produced physics, chemistry, and evolutionary
biology.
The most important innovation in linguistics of this time was the development of a method for the comparison of
languages. One of the first lessons learned is that it is easy to find false cognates: words that look similar in two
languages but are not related, due to a systematic change that has taken place in phonology of one of both of the
languages. The comparative method provided a way in which comparisons could be made between languages in a
more systematic and less error-prone manner. This research agenda also suggested the need for systems of
transcribing sound that were not language-specific, so that such comparisons could be made; this need gave rise to
phonetic alphabets culminating in the International Phonetic Alphabet (IPA).
Another intellectual current arising especially in the 18th century was the development of prescriptive grammars.
Grammar until this time had mainly been seen as something that was a property of Latin and Greek. It was in the
18th century that European scholars began to write grammars and dictionaries for ‘vulgar’ languages such as German,
French, and even English.
All of these developments changed the way people thought about language. An early way of thinking about language
was through a biological metaphor, in which languages were seen as organisms. Another view held that language
was the ‘personal property’ of an individual. This argument was debated vigorously, with the opposing view being
that language is a community object.
In the late 19th century, the Neogrammarians radially rethought the relationship of language and individual. They
postulated that it is not individual words that change in a language. Rather, linguists should be searching for regular
and predictable laws of sound change, just as physicists search for laws of nature. That is, languages change through
individual sounds (or categories of sound) rather than word by word. The Neogrammarians were (and still are)
central to variationist work because they began to look for ‘laws’ of sound change. They began to describe changes
in ways that look similar to phonetic and phonological changes as they are described in modern historical linguistics.
The first revolution in linguistic thinking in the 20 th century was started by Saussure and especially by the notion of
the phoneme. Phonemes are defined by their structural relationship to other phonemes in the linguistic system, and
not by their phonetic character. In this view, the significance of /p/ in ‘pin’ is not the phonetic character combined
with the other phonetic characters of the word, but its difference from other phonemes such as /b/ or /d/, and this
can easily be seen by the fact that replacing /p/ with these other two consonants produces words with completely
different meanings. A structural relationship is one of difference that produces meaning, crucially within a particular
linguistic system.
While Saussure worked in Geneva, in America the field of anthropology was beginning to develop. Franz Boas saw
the project of linguistics as part of anthropology and as challenging the view that American Indians were somehow
inferior to Anglo-Americans. One of his tools in this endeavour was to describe American Indian languages in a
rigorous way. In addition, Boas stressed the ‘unconscious’ nature of language, and aligned this character with
‘culture’, since both have categories that develop organically rather than through conscious thought and discussion.
Edward Sapir developed Boas’ ideas and expanded the anthropological linguistic tradition of description. This current
became an important one that in the later 20th century influenced sociolinguistics considerably, both because of its
focus on descriptive methods and because in this tradition language was always seen as a cultural rather than
individual property.
The tension between language as an abstract object and as the property of an individual, however, is still present.
Weinreich, Labov and Herzog (henceforth WLH) spent a considerable amount of time engaging with the theories of
change outlined by Hermann Paul. Paul’s view of linguistic change rested on a concept of language as something
that belonged to an individual: he argued that the ‘idiolect’ is the proper object of linguistic study. A dialect (or
language) for Paul is a collection – an average – of idiolects. Language changes when idiolects change in parallel, and
dialects split from one another when groups of idiolects diverge. These changes are explained by contact with other
dialects and accommodation by speakers to that new dialect. WLH spend considerable space refuting this view in
favour of a view of a dialect as a community object. They argue that Paul’s view cannot account for the reality of
language change, and indeed can account only for phonetic changed and not for change in the case of lexical
replacement.
WLH delve deeply into many other approaches to change. It is worth understanding the innovations in the study of
language change established by WLH. They focused on the heterogeneous nature of language, as opposed to the
homogeneous. Underlying this focus is an understanding of language as existing in a community of individuals, rather
than completely and perfectly in one of those individuals. The main difference between WLH’s view and that of
previous theories is that language or language variety (dialect, etc.) is not the averaging of individuals, or even
something locatable in a single individual’s mind, but rather an external, community ‘object’. In this view, individual
variation is not to be averaged, but rather seen as the pattern of the variety: variation is part of the very description
of a language.
WLH argue that ‘the identification of structuredness with homogeneity’ is one of the main conceptual problems
preventing linguistic theories from developing an adequate account of linguistic change. Really, there is no
observational or scientific reason to assume that language is homogeneous (whether in a community or in
someone’s mind).
One of the most obvious examples of this ideology of homogeneity is the focus of structuralist phonology on
exceptionless (called categorical), predictable rules or constrains. That is, in describing a language, the focus is on
predicting categorically the phonetic character of phonemes. In this view, if something is not categorically
predictable – i.e., if the rule does not apply always, - then it is ‘free’ or without order.
So WLH introduced the term ‘orderly heterogeneity’, which addresses the fact that earlier theories had viewed any
type of heterogeneity as disorderly. Of course, we see heterogeneous order all the time in almost all human
practices. For example, when do you get out of bed every morning? At dawn? In fact, it would silly to suggest that
everyone gets up exactly at dawn, and we wouldn’t argue that everyone has this as the necessary ideal. But this
doesn’t mean that there isn’t a predictable, generalisable structure to the way a community, a culture, or an
individual wakes up each day; in fact, there is an approximately four-hour window in which most people wake, and
the waking times of those outside those patterns can be accounted for in other ways. It is not hard to think of human
knowledge and behaviour as being heterogeneous, or variable, but still possessing predictable structure.
One of the earliest and best-known variation studies is Labov’s investigation into linguistic change in New York City.
In New York the /r/ is not always pronounced. Labov studied whether there was a pattern in who pronounced the
/r/ and when. Part of his method consisted of going to three different department stores and asking where some
item was located in the store. Both the pronunciations of individual speakers and pronunciations across speakers
were variable. By looking at the historical record and by comparing older and younger speakers, Labov determined
that the r-full pronunciation was increasing in New York.
We can use this example to see that there are two dimensions of orderly heterogeneity.
1. First, speakers behave differently (heterogeneity), but which speakers and which utterances are more r-full
will be statistically predictable (orderliness). Labov found that salespeople in all stores had variable
pronunciations, but that those who worked in the high-end store used the r-full pronunciation more than
those in the other two stores.
2. Second, pronunciation will be affected by linguistic factors: in this case, the pronunciation of /r/ will be
statistically predictable on the basis of some other linguistic fact, such as whether the /r/ is followed by a
consonant in the syllable (this is orderliness).
So Labov found heterogeneity, or variation, both across individuals and within individuals, and this variation was not
random, but patterned in statistically predictable ways. Orderly heterogeneity allows us to see a change such as in
/r/ not as an abrupt change but as a more gradual and organised affair. This reconceptualization also sets up new
questions, such as whether there are universal patterns of variability. WLH articulate a set of problems that a theory
of linguistic change must answer, and one riddle that may be unanswerable:
1. The constraints problem, or that of ‘the set of possible changes and possible conditions for changes which
can take place in a structure of a given type’. In what linguistic and social conditions are certain changes
likely or unlikely?
2. The transition problem: What intervening stages ‘can be observed, or must be posited, between any two
forms of a language defined for a language community at different times?’ The New York example tells us
that there needs to be some kind of variability in the transition. Theoretically, we will want to try to predict
what these stages will look like for all changes, or perhaps for different kinds of changes. Finally, the
transition problem encompasses the regularity question: do changes move regularly as changes to
phonemes, or do they move at different rates from word to word?
3. The embedding problem: what other changes are associated with the given changes in a manner that
cannot be attributed to chance? This is the problem on which the vast majority of variation studies have
focused. The theoretical goal in this area is thus to find regularities such that we can predict the ways that
changes are likely to expand through a community and through a language.
4. The evaluation problem: ‘How can the observed changes be evaluated – in terms of their effects upon
linguistic structure, upon communicative efficiency, and on the wide range of nonrepresentical factors
involved in speaking? Theoretically, we want to look for predictable, repeated patterns of how such changes
are usually evaluated in the community, and whether there is agreement or disagreement on how and even
whether to evaluate the new and old ways of speaking.
5. The actuation riddle: How and why do changes begin? This is a riddle because changes probably begin
unobserved, making it difficult to identify their start.
Once we take variability to be a basic design feature of language, we then need a way of theoretically, or at least
heuristically, representing it. This representation has been accomplished through the notion of the linguistic
variable. The essential idea is that there is one, isolable linguistic feature that carries meaning, such as the
pronunciation of a phoneme like /r/, and the community has more than one way of representing it in language.
WLH end their essay with some general principles for the study of language change:
1. ‘Language change is not random drift proceeding from inherent variation’. However, there seems to be
more directionality in language change than would be predicted if change were random and unstructured.
2. ‘The association between structure and homogeneity is an illusion’.
3. Variability is a ‘natural’, even stable state for language, but if language does undergo change, we are
logically going to see some variability.
4. ‘The generalization of a change is neither uniform nor instantaneous’. Changes happen faster in some parts
of the language and for some speakers, and there will be many speakers who show variability in the
transition.
5. ‘Grammars of change are grammars of speech communities’. It is one of the most difficult principles for
students to understand. We cannot understand language change by assuming a single, ideal individual who
represents the only grammar during the change. Rather, the grammars we represent are those of a speech
community. Labov described /r/ in New York City, not in a particular New Yorker. Of course, speakers do
have linguistic knowledge, and the problem of how to represent variable knowledge is an important one.
6. ‘Linguistic change is transmitted within the community as a whole; it is not confined to discrete steps in the
family’. There are no big jumps from mother to son, for example, and the change is not restricted to families
but affects – and is moved forwards by – the entire community.
7. ‘Linguistic and social factors closely interrelate’. Linguistic change does not happen in every speaker in the
same way, but rather certain types of speakers will adopt the change before others. The linguistic and the
social constraints (or factors) affecting a change are to some extent independent, but they will interact.
Thus, almost everyone in New York will show the effect of a following consonant, but their overall rate of
/r/-fulness will vary as well.
We will include a third type of factor. These are cognitive factors, and they constrain what speakers cognitively
perceive and produce. One of the central problems in variationist linguistics is to determine exactly how different
factors relate to one another. For example, does the following environment for /r/ in New York work the same for
everyone?
All three of the major constraints (structural, social and cognitive) act on language simultaneously, and it is the
tension among them that ends up driving and determining the direction of changes.
The usual definition of a linguistic variable is that there is two or more ways of saying the same thing, where the
‘same thing’ refers to what is denoted by an utterance. For example, the alternation between walking with [ɪŋ] and
walkin’ [ɪn] shows that there are (at least) two ways of uttering {ing} morpheme without changing the central
meaning of the word.
This notion was introduced by Labov in his article on the island of Martha’s Vineyard. Labov noted that some
islanders had a distinctive way of pronouncing /aʊ/ and /aɪ/ as in down and bite. He showed that these phonemes
were undergoing a change towards centralisation of the first element of the diphthong. Labov also introduced the
notation convention of using parentheses for variables, so when discussing the Martha’s Vineyard variables he used
(aʊ) and (aɪ). This notation distinguishes them from the phonemes /aʊ/ and /aɪ/, and also still allows a discussion of
their phonetic characteristics with square brackets, so that the variable (aʊ) might have the phonetic variants [əʊ],
[aʊ], or [æʊ]. For the morpheme {ing}, the variable is (ing), with variants of [ɪŋ] and [ɪn].
Linguistic variables can occur at any linguistic level: phonological, morphological, lexical, syntactic, pragmatic,
discoursal, and suprasegmental. The trick is to define carefully the part of the variable that ‘stays the same’. The
most important issue is to be able to identify the variable context. That is, to be able to find every place that a
speaker must make the choice we are interested in.
Phonological
The two examples given so far are examples of phonological variables. For the Martha’s Vineyard (aʊ) variable, the
variable context is every instance in which English phonology requires this phoneme, no matter how it is said. For
(ing), it is a little more complex. It is tempting simply to identify any -ing spelled cluster as the variable context. But
this definition would be too broad, because there are certain instances in which there is no option of using the [ɪn]
variant, such in words like king. If the variable context is not completely clear, it is best to define it broadly and then
exclude contexts that do not show any variability.
Morphological
The variable context of morphological variables is usually identified by the morpheme’s grammatical function, such
as tense or aspect marking on verbs, or plural marking on elements of a noun phrase. So, for example we might say
that plural agreements on nouns in Spanish is a variable.
Lexical
Lexical variables are fairly straightforward, as long as we can show that the two variants – such as the choice between
soda and pop for a carbonated drink in American English – refer to the same entity.
Syntactic
Variables at the level of syntax are some of the most difficult to define, because it is not always clear whether or not
two variants are equivalent at some abstract level. Nevertheless, speakers generally have some choice in what clause
structure to use in their utterances. For example, there is alternation in English between passive and active
constructions.
Pragmatics/Discourse
The most difficult type of variable to determine is at the pragmatic level, in which an abstract pragmatic function is
taken as the variable. A choice of word order assumes that there are functional reasons for the choice. In this sense,
the choice is assumed not to be equivalent in terms of meaning, but the choice is nevertheless important.
Suprasegmental
The variables are most often pitch contours or intonation, and very little work has been done on them.
There are therefore choices in how to say something at every level of language, and every possible choice can be
considered to be a variable. The difficulty is in determining exactly what possible choices exist. The way in which
variable contexts are defined is one of the main differences among variables at different levels of language.
Most of the variables outlined above constitute discrete variants, enabling a clear choice between one form or
another: [ɪn] or [ɪŋ], soda or pop, passive or active. However, it is possible in the phonetic realm to have continuous
variables, with an unlimited number of choices on a continuous scale. In the Martha’s Vineyard study, Labov had
four levels of centralisation for each vowel, which he coded auditorily.
Whether we have a discrete or continuous variable is important in choosing a statistical method for analysis. The
choice of variants depends on the analyst: the extent to which variants are imposed on the data, that is, any category
created by the analyst, and may or may not be salient for the user of the variety being analysed.
A further important distinction is between those discrete variables that have only two variants and those that have
many. Here again the imposition of analytical choices makes a difference in how the data are understood.
In sum, we have two different ways to characterise linguistic variables: based on the level of language, and on the
discreteness of the variable. Given the variability of what counts as a variable, we will simply say that a linguistic
variable is a choice or option about speaking in a speech community. Notice that this broad definition is based on
the speech community, which is important because something that is a variable in one speech community may not
exist in another. For example, the deletion of word-final /s/ is variable in some Spanish speech communities, but in
others it may not be so at all.
The variable rule is an important concept in the development of variation theory. Although the concept is no longer
generally used, it must be grasped in order to understand what objections to the notion of variable have already
been addressed, and also in order to appreciate more fully what the concept of the variable is.
The first discussion of a linguistic variable was by Labov in 1963, but it wasn’t until 1969 that he devised the first
formal variable rule. This article was on the realisation (or deletion) of the English copula in African American
Vernacular English (AAVE) and mainstream American English. Labov’s concern in this article is to show the regularity
with which the copular is deleted in AAVE in sentences:
The ‘Ø’ indicates that the copula is (or the contraction -s) has been deleted in this utterance. Labov formulates
generative morphophonological rules in the fashion of the time. These rules are ordered and apply at different rates.
This rule states that a schwa is deleted variably. The rule applies variably in certain contexts.
The important point here is that ordered heterogeneity was being inserted into formal linguistic theory. The fact
that the rule is variable is a representation of heterogeneity: the rule applies sometimes, and sometimes not. But
it does not apply randomly, and this is indicated by the ordering of the different constraints. With the variable rule
concept, Labov thus provides a proposal for how to implement the claims made by WLH in formal linguistic theory.
The things that affect the rate at which a variant appears we have been referring to in a non-technical way as
constraints; the same terminology is used by Labov. However, later variationist terminology, which we will use from
now on, discusses constraints as factors or factor groups.
The way this ordering was represented in early variationist work was through a cross-product display, which shows
all the possible combinations of factors and the rates of variation for each combination. Fasold provides a particularly
good example of such a representation in his discussion of the variable /t/ or /d/ deletion in English. Fasold provides
a cross-product display based on data from Wolfram, a study that focused only on the deletion of /d/. Fasold includes
three constraints: whether or not the following word begins with a vowel, whether the syllable with the coronal
segment is stressed or not, and whether the final segment is part of a suffix.
There are two problems with this kind of schematisation. The first is that some combinations of constraints may not
be possible. A further complication arises if there is a three-way option rather than a binary one.
A more sophisticated method and representation was developed by Cedergren and Sankoff. They modified the
variable rule notation and provided a statistical method for determining the probabilities which should be assigned
to constraints by using angled brackets.
The method they use, called logistic regression, produces the probability that a rule will apply for a particular factor
independent of the other factors. Labov argued that factors in variable rules should be ordered, and that the final
frequency of the rule application was obtained by adding up the frequency for each factor, as seen in the cross-
product display. Cedergren and Sankoff argued that we should calculate the probability that a rule will apply given
a particular constraint. Moreover, these probabilities should be multiplied to determine the final rate. One important
consequence of operating in this manner is that the strength of the constraint would be represented by the
magnitude of the probability. Probabilities range from 1 to 0; 1 means that a rule will always be successful, while 0
means that it will always fail. A probability of 0.5 means that a factor has no effect – the rule applies randomly. So
stronger constraints will have a value closer to 0 or 1, weak constraints closer to 0.5.
Once we get the probabilities from the statistical program, they can be multiplied together to arrive at the predicted
frequency of the variable. Moreover, each factor is assumed to be independent.
We assume that the following segment always contributes the same probability no matter what is happening in the
other factor groups, and each operates independently of the overall tendency if no factors are taken into account.
For the most part in these studies, variable rules have been abandoned in favour of simply noting the probability
that one variant will surface.
Another criticism that has been levelled at the original definition of variable is that it privileges a particular ideology
of language with a limited view of what counts as meaning: a way of thinking and talking about language that sees
the only function of language as a system of referring unproblematically to ideas and things.
Symbolic meaning holds when a linguistic form unproblematically ‘means’ or stands for an abstract idea or
thing. Thus, three in English refers to a number, and in another language it will be arbitrarily represented differently
(drei, san, troi, tres, etc.). So if someone in an English-speaking speech community sometimes says [tri] for three, we
can argue that the /Ɵ/-/t/ alternation is a variable because the symbolic meaning is the same.
Indexical meaning. An index is a sign which takes its meaning from contextual associations. For example, a word
that is always used by a particular person may take on a ‘meaning’ that indicated this person is speaking. So if
someone other than the person says it, the person can be seen as ‘voicing’ the person who usually does. For example,
if Uncle Eric always says ‘by jingo’, and I start an utterance with that phrase when speaking to someone who knows
him, then I am indexing that I am speaking as if I were Uncle Eric.
One could argue that any choice or option in language is likely to have some kind of indexical meaning. If we consider
meaning in this way, then the classic definition of ‘variable’ is again problematic. However, we can respond by using
our expanded definition of ‘variable’, and then argue that one of our interests is in discovering what meaning
differences hold for the various choices.
Other criticism focuses not on the variable itself, but on the claim that variables and the frequencies of their variants
represent a consistent community pattern or grammar, rather than an average of individuals. Bickerton points out
that it is at least theoretically possible that a variable pattern in a community could arise because there are
individuals categorically using one variant or the other. When the group is averaged overall, it may look as though
the entire community is variable, but it is actually the case that there are two groups of categorical speakers.
The most careful and complete refutation of the criticism was produced by Guy, using again the CSD (coronal stop
deletion) variable. He showed first that for his data, there was ordering for three factor groups:
Guy used the Varbrul program to calculate probabilities for each individual in his study, and then compared the
relative ranking of factors in factor groups for each individual to that of the overall community.
There is evidence that speakers’ patterns of variation reflect those of their speech communities, because the
differences were directly related to speech community membership, and not random individual differences.
The introduction of the idea of the variable rule had a profound effect on the way variables are conceived of. It made
it possible to formulate the variable and its variants outlining alternative realisations of the variable. The variable
rule introduced ideas from generative linguistics into the mix, by suggesting that variation was the outcome of a
process or rule rather than of choices among alternatives. This imposes a more restricted set of alternatives on the
variable, because they need to represent ‘underlying forms’ and the output of a rule. This view has meant that the
majority of work in variation has had variables that are limited to two variants. Even when three or more possibilities
are recorded in the coding, two or more variants are usually combined in order to produce a binary variable. The
early development of the variable rule thus led to a statistical procedure – Varbrul – which is expressly designed to
handle binary variables, and in fact forces one to create a binary variable.
These criticisms of the notion of the variable have led linguists to a more flexible and nuanced understanding of
optionality and choice in language and linguistic change. The requirement for identical meanings has been relaxed,
and new statistical methods have been developed to handle situations in which one speaker may have undue
influence on an overall data set, and situations in which there are more than two variants available to speakers.
We well conclude by mentioning two criticisms that are aimed at the dangers in taking the means of variationist
analysis to be the ends. Both have to do with the relationship between the methods of variationist analysis and
conclusions that can be reached from those methods.
Variationist linguistics is a deliberately data-driven and empirical method, but there are issues of interpretation at
many points. Two dangers to watch out for are the tyranny of correlation and the atomisation of category.
The method of variationist linguistics is based on correlating a linguistic variable with other factors. What
comes next, however, is interpretation. The tyranny of correlation comes about when this interpretation attributes
too much to the correlation, in terms of either meaning or significance. For example, the correlation between being
a Pittsburgher and using /aʊ/ -monophthgisation suggest that (aʊ) has a ‘Pittsburgher’ meaning. This understanding
of the variable is simplistic at best, because when asked about this way of speaking, speakers diverged considerably
on their identification of the variant and the Pittsburgh way of speaking. This problem is only solvable by being
linguistically and socially thoughtful in our analyses, and by always being sceptical of our results.
The atomisation of category is brought about when we atomise different aspects of language, identity, and
community into small, component parts, the better to correlate with the linguistic variable. There is an almost
infinite number of possible factors potentially influencing a linguistic variable, and we choose among them. Thus we
have already delimited the possibilities of describing the behaviour of a linguistic variable in a speech community by
creating these categories for analysis. Moreover, we assume the factors don’t influence one another separately from
the linguistic variable. However, this assumption of independence is simple that: an assumption. It is entirely likely
that some linguistic variable and factor groups may interact. These problems can be met with thoughtful methods
and careful, evidence-based interpretations.
UNIT 2 GLOSSARY
Discrete variants > these are variants in which it is enabled a clear choice between one form or another (for instance
[ɪŋ] and [ɪn]). This is the most common way of conceiving linguistic variables.
Idiolects > It is an individual's distinctive and unique use of language, including speech. This unique usage
encompasses vocabulary, grammar and pronunciation. This differs from a dialect, a common set of linguistic
characteristics shared among some group of people
Indexical meaning > When a sign takes its meaning from contextual associations. An index assumes that a linguistic
form is used in a particular situation and from a particular person.
Language constraints > They are a set of possible changes and possible conditions for changes which can take place
in a structure of a given type.
Linguistic variable > It is a choice or option about speaking in a speech community. Linguistic variables can be
characterized in two different ways: based on the level of language (phonology, morphology, etc.) and the
discreteness of the variants.
Neogrammarians > They were a group of German philologists who developed the theory that language change is
systematic and regular, and governed by a set of rules. According to them, language changes through individual
sounds rather than word by word. They established the Principle of regularity, which was based on two fundamental
principles: 1) Phonological change takes place according to mechanical laws that have no exceptions. 2) Importance
of analogy in the creation of new forms. Their attention focused mainly on diachrony.
Orderly heterogeneity in language > This term addresses the fact that earlier theories had viewed any type of
heterogeneity as disorderly. It refers to the fact that speakers of a language have many choices in how to assemble
any utterance, but that these choices fall under a predictable pattern in terms of the linguistic system and social
factors. Labov found heterogeneity both across individuals and within individuals, but this variation was not random.
Methodology is an integral part of variationist linguistics. Variationists take the view that methodology is inherent
to theory and to the understanding of language. Basically, methods are in the service of the questions we are asking
about language, so these questions shape the methodology. Since we are searching for answers about speech
communities and not individuals, we must analyse language that is representative of the community and not just an
individual. By focusing on the speech community, we will want to record speech from many members of the
community in order to understand the language of that community.
First, we must identify a variable or variables to focus on. This might be done on the basis of known variation in the
community, but it also might be the focus of a preliminary study. We also want to determine whether we think the
variable is undergoing change or whether it is stable. Recall that there were five main problems outlined by WLH,
which can be turned into the questions that might drive a research project:
1. The constraints problem: What are the linguistic, or internal, constraints of the variable? What linguistic
conditions have led to the change?
2. The transition problem: If the variable is changing, how is it spreading through the linguistic system and
through the speech community? At what stage is the change and why? Who adopts the change first and
who adopts it last? How does one generation move the change forward?
3. The embedding problem: How is the variable patterning in the community and the language?
4. The evaluation problem: Does the change affect the communicative efficacy of the variety? How do
speakers evaluate other speakers who use one variant more than another? Why do they evaluate speakers
this way? What effect do these evaluations have on the change, if the variable is undergoing change?
5. The actuation riddle: How and why did the change begin?
For the most part, the actuation riddle has been set aside because the methodology required to study it does not
exist. however, there are two likely mechanisms: one is that there is always some variation in a language, especially
in phonology, and so one generation could easily begin to shift that variation to the point that it starts a change.
Alternatively, and more importantly, speakers use language creatively, using linguistic forms in new ways, and this
expressive creativity may lead to many changes. Most probably the answer is a mix of these two mechanisms, and
we can investigate the actuation riddle indirectly by understanding variation in language that doesn’t change, and
also how speakers use language creatively in interaction. The actuation riddle is of course related to every other
problem since actuation precedes everything else.
Given those issues with the actuation riddle, let’s turn our attention to the other four problems and their relationship
to methods.
The constraints problem suggests that we need to know a lot about the entire variety we are concerned with. That
is, we need to record many aspects of language that might affect the variable. How we decide on what constraints
to record depends on our knowledge of linguistic theory and on likely effects that have been previously documented
for the variable or for similar variables. The constraints we find are therefore intimately tied up with our methods:
what kinds of recordings we have made, and what kinds of measurements we have made of the possible constraints.
Real-world limits on resources (especially time) usually require that we select only those possible constraints that
our models of language tell us are most likely to have a significant effect on the variation.
The transition problem suggests that we want to try to find linguistic changes in progress, and be able to identify
which linguistic variables are involved in change and which are variable but not really changing. Of course, the most
obvious way to do this is to record speakers from the speech community at different times. However, most studies
do not have the luxury of the several years or decades needed to detect changes that take place in generational
time. When this kind of study is impractical, it is sometimes possible to find records of earlier studies of the speech
community, even if performed using different methods. For studies of syntax, real-time historical studies of texts
have yielded important results. Syntactic changes seem to happen over a much longer time span than phonological
changes, but they have the advantage that texts of many languages exist over hundreds of years.
The studies that compare speech recorded at different times are called real-time studies, and give us the most
information about changes in a speech community. However, the vast majority of variationist work is done with
apparent-time studies, which are apparent because they compare speakers of different ages or age groups recorded
at a single time. If there are significant differences in the use of a variable from the older to the younger generation,
then it is hypothesised that we are seeing a change in progress in the direction of the younger generation’s usage.
Of course, it is entirely possible that as speakers age they change the way they talk, a pattern known as age grading.
However, in general these changes reflect generational changes, or changes in the speech community, rather than
age grading, especially for phonological systems. Different kinds of variables seem to be more or less susceptible to
age grading. In addition, age grading shows that older speakers change in the direction of overall change in the
community.
In addressing the transition problem, ideally, we triangulate how a transition is happening by using both apparent-
time and real-time evidence. However, we can also be fairly confident that if we find differences in variable usage
by age, we are finding a change in progress and the study can inform our understanding of the transition problem.
The embedding problem focuses on how the change is embedded in, or moves through, the speech community. The
central issue in addressing this problem is making sure that our data are representative of the speech community.
We have to be careful about which speakers we record, and to keep track of the identities and practices of those we
record and their positions in the community. For example, we need to have speakers of different ages in order to
make claims about change in apparent time. More important for the embedding problem is that some speakers of
the same generation adopt linguistic change before others. Our goal is to discover these patterns and they try to
find a reason why they exist. So, we need to know also whether, for example, women adopt changes before men,
or upper-class speakers adopt them before lower-class speakers, etc. This problem, then, is one of sampling a
particular speech community in a way that represents the important types of person in it.
The final problem is the evaluation problem. This problem focuses on how people evaluate other speakers who use
different variants we are interested in. This problem is thus one that points to more experimental methods of
presenting language to members of the speech community and measuring their reaction to it by various means.
ETHICAL LINGUISTICS
From very early on, sociolinguists have been concerned with helping the people who agree to provide data. This
concern begins with permission to record someone’s speech. We always obtain informed consent for recordings
before those recordings are made, usually in written form. Ethics panels and internal review boards (IRBs) will need
to provide approval as well, and will insist on informed consent. Informed consent means that speakers need to
know more than simple fact that they are being recorded. They should also have some idea of what we will do with
that speech. However, it is usually best to not be too specific, for two reasons. First, speakers are probably not
interested in the intricate details of the project. Second, if we tell them very specific things like what variable we are
interested in, then we have called attention to it and speakers may adjust their use of the variable whether they
mean to or not.
Sociolinguists, however, have long argued that our ethics can go further. Some sociolinguists argue that in fact we
should have a net positive effect on the communities we study. There are three principles that Wolfram outlines:
• Principle of error correction: ‘A scientist who becomes aware of a widespread idea or social practice with
important consequences that is invalidated by his own data is obligated to bring this error to the attention
of the widest possible audience’.
• Principle of debt incurred: ‘An investigator who has obtained linguistic data from members of a speech
community has the obligation to use the knowledge based on that data for the benefit of that community,
when it has need of it’.
• Principle of linguistic gratuity: ‘Investigators who have obtained linguistic data from members of a speech
community should actively pursue ways in which they can return linguistic favours to the community’.
1. A speech community or communities to study is or are identified, and at least one linguistic variable, is this
can be identified in advance.
2. If the researcher is not already familiar with the community, preliminary ethnography and study should be
performed, focusing on understanding how members of the community understand its boundaries and
what identities and social practices are likely to correlate with variation in the linguistic variable.
3. The parameters of the speech community are defined, and a strategy for recording speech samples,
attitudes, and social practices of members of the community is designed.
4. The data are collected.
5. Linguistic variables and factors are coded.
6. Significant factors affecting the linguistic variables are discovered using statistical tests.
7. Other data are applied to explain the correlational patterns that are found. Follow-up studies might include
further interviews asking about speech.
This process is pretty straightforward, but the outline above hides many potential complications and problems that
require further consideration.
The first problem is to decide what counts as a speech community, and specifically what counts as the speech
community for a particular study. Let us approach this problem by focusing on the concepts that recur in the
definitions and discussions of speech community.
There are two central ideas that recur in many definitions. First is the notion of some kind of shared linguistic norm.
These shared norms can refer to similar ways of producing speech, but also often refer to language more generally.
The shared norms thus not entail everyone speaking alike, but could mean that speakers produce language
differently and at the same time evaluate speech in a similar manner.
The second central idea is that there are some kinds of communication lines that link the members of the
community. In other words, a speech community is a group of people who talk to each other more than other
possible interactants, whether for social or physical reasons. There are many problems with both of these central
recurring ideas.
There are many practical issues that add to the problems with the speech community idea. The first is that
sociolinguists have tended to use geographical boundaries of varying sizes and precision to define speech
communities and their members. However, in none of the definitions of speech community do we find an essential
focus on residence or provenance for inclusion or exclusion in a speech community. Moreover, while space can be
defined by precise coordinates and boundaries, place – how people think about their physical surroundings – may
not be as precise and in fact different speakers in the same space may think of place differently. Another convenient
way to label a speech community is by identity category such as race, ethnicity, class, or gender. Here again we have
the problem that some convenient external label is applied, which may not actually match the speech community.
From the above discussion, it should be clear that any operational definition that outlines criteria for membership
will always be imperfect. The important question is: how do we know before we perform a study who should be
included in the speech community? We want to know this because if we decide what speakers are in or out of the
community after we interview them, then we could bias our study. The practice in most studies is to decide in terms
of geographical residence, identity category, or other non-speech property of individuals, informed by some
knowledge of the variety in question ahead of time. There are methods for determining how much speakers consider
themselves to be from the same community or region. For example, Zelinksy, a geographer, mapped regions of the
United States by recording names used by business in different cities. One could also survey speakers as to their ties
to a particular community, either by frequency of interaction, by residence or by asking whether or not a speaker
consider him/herself to be of one particular place.
Such methods avoid, for the most part, the problem of circularity. The danger is that one might argue that a speaker
is not part of the speech community because he or she has a different pattern of variable use from others originally
in the speech community. If the speech community concept is to be useful at all, we need either to avoid this
circularity or to build it into the methods. We avoid it by using objective methods such as the regional naming
discussed above. We build it in by taking a wide view of the speech community and declaring that one goal of our
research is to discover the extent to which a variant is used geographically or socially.
The theoretical issues surrounding the speech community are still unsolved in sociolinguistics. At the same time, it
remains a central concept in variationist sociolinguistics, and this status is likely to continue as well. The best advice
that can be given in the practical case of designing a research project is to think carefully about how we are going to
use the speech community concept, and to be explicit about the assumptions we are making about a speech
community in each study.
Here are some questions about speech communities to answer when beginning a variationist project:
• What are the precise characteristics of a person who is in (or out of) this speech community?
• Why are these the criteria for membership?
• What are the benefits and drawbacks of defining the speech community in this way? How do the benefits
outweigh the drawbacks?
• If we are discovering the speech community, what criteria will count for boing in or out of it (ways of
speaking, self-report of belonging, evaluative norms of language, etc.)?
There are other concepts which, like speech community, serve to group people together for sociolinguistic study.
• Social networks are groups of people who have sort of social tie to one another (as family, friends, co-
workers, neighbours, etc.). It should be obvious that this way of looking at a group of people is similar to
the ‘mutual concept’ criterion sometimes posited for a speech community. One could identify a network,
or even a single speaker’s network, as the group from which to sample. In fact, one method often used by
sociolinguists is essentially a network sampling: We start with a few individuals whom we already have met
in the community, or introduce ourselves, and then ask them to introduce us to friends, and keep asking
for introductions. This ‘snowball’ allows us to follow several networks within a particular community.
• A community of practice is a group of people that is organised around some practice or activity, such as
chess or book club. The practice can be explicitly socially oriented, so that a group of people who dress
similarly or have similar musical tastes can form a community of practice. Such a community is similar to a
network, with the difference that the community of practice is organised around the common practice,
rather than around individual ties to other individuals.
From a methodological point of view, these concepts – speech community, network, and community of practice –
provide us with different ways of identifying speakers whom we would like to record. Ideally, we would record
speech from every single individual in our speech community, but of course this is impossible. This group of everyone
is called the population by statisticians. Since we cannot record everyone, we have to record a small group of the
population, called a sample, that is as close as possible to the population in every characteristic. The most statistically
rigorous is the random sample, in which a certain number of people are chosen from a population (speech
community) completely at random. However, in sociolinguistic studies this type of sampling is almost impossible to
achieve, because the methods used to contact speakers are likely to bias the sample in one direction. Most studies
therefore use a judgement sample, which is as random as possible but also determines ahead of time how many
people of certain characteristics should be included, often on the basis of the proportion of that characteristics in
the population as a whole. For example, the sample might be balanced in terms of age group, gender, class, ethnicity,
or any other social aspect that the researcher suspects may affect the use of the linguistic variable under study. One
problem with these samples is that they can bias the sample towards one group if the quota is not correct. In
addition, they may extend the time it takes to complete the recordings because the last few speakers may take an
inordinate amount of time to track down. These problems are not as prevalent for snowball sampling.
The key is to make sure that the kinds of research questions one is asking match the methods. For example, are we
producing a survey of a speech community never studied? In such a case a random sample might be desirable. If on
the other hand we are studying the meanings attached to variants in interaction, then a snowball sample or even a
small community of practice might be the best choice.
Finally, we come to the question of sample size. How many interviews should I do? The answer is: ‘It depends’. Note
that statistics are generally based on the tokens of the variable and not on, for example, on the average of each
speaker. Thirty people are generally assumed to provide a decent sample size, although for really robust statistic
reliability one wants more. The most important issue for this question, however, is what, and how many, groups of
speakers you want. If you want to compare, for example, four age groups, two genders, and four social classes, you
have to have a category for each combination of age, gender or class. Each interaction of all three is called a cell,
since the design can be represented as follows.
The question then becomes how many speakers should be in each cell. Some researches opt for five. The sample
does not need to be balanced, but we should try to have a minimum for each cell.
When we record speech, there are several characteristics we might wish for. First, we might by trying to record the
speaker’s most ‘natural’ way of speaking. Second, we might be striving to record with the most fidelity possible.
Finally, we want to record enough speech so that we have a large number of tokens of the variable we wish to study.
These goals are sometimes at odds, and can also be at odds with our ethical goals. For example, ‘natural’ speech is
more likely to take place when someone doesn’t know they are being recorded, but for reasons of fidelity and ethics,
the speaker needs to know we are recording.
It is for this reason that most variationist studies rely on the sociolinguistic interview, in which a researcher asks
questions of a speaker and records the answers. we use various interactional ‘tricks’ to attempt to make the
interview as close as possible to a ‘normal’ casual conversation, but the fact remains that the interview is not the
type of speech event that speakers engage in during their day-to-day activities. Other methods are possible, but
there are always trade-offs.
The sociolinguistic interview has therefore been thought to represent the best compromise among ‘naturalness’,
fidelity, and amount of speech. Researchers have developed quite sophisticated techniques that create a much more
conversation like atmosphere for the interview. Nevertheless, our findings based on such interviews should always
be viewed with the right amount of caution, because there are always factors that affect variability that we are not
controlling for.
In designing and listening to interviews, we need to realise that a speaker will always be talking in relation to the
person they are talking to. Therefore, even though we hold the interview constant, the differences we find in
linguistic variable use among interviewees may not be differences in their ‘vernaculars’, but rather in how they relate
to the interviewer.
The technical details of recording have changed considerably since Labov made his recordings on Martha’s Vineyard.
Modern equipment makes it possible to record speech on one’s mobile phone, whereas before such recording
required bulky reel-to-reel equipment. However, we must not lose sight of the fidelity requirement in the fog of
convenience. Even if one does not plan to perform sensitive phonetic analyses of speech, it is important to get the
best fidelity possible, if only because one might later on wish to perform that phonetic analysis. The best way to
record a speaker is by using a small microphone which clips on to clothing, usually at the collar or lapel.
Most recording is now digital, and good-quality digital records can be purchased by a reasonable price. For a few
years there were some critics of digital recording, arguing that because digital recorders only sample the sound, we
do not know what we are losing. But if the sampling rate is high enough, there is no noticeable loss of data. Most
laptop computers now come with recording capability, but this is not the first choice for field recording.
Always familiarise yourself with your equipment before you begin research; perform several ‘mock’ interviews with
friends in which you set up your equipment. The more you have to fiddle with equipment when doing interviews,
the more likely you will be to call attention to the recorded aspect of the interview.
On a final practical note, be very careful to keep track of all recordings, and to make more than one backup, kept in
at least two places. Do not think you are immune from loss; while you may have never lost recordings due to theft,
damage, or computer malfunction, you may have known too many others who have lost data for all these reasons.
Label the file names logically, and keep a catalogue of what is in each file.
CODING VARIABLES
Coding the speech means measuring and recording each token of the linguistic variable and recording any social or
linguistic variables that go with that linguistic variable.
The first decision in coding is to determine what counts as the variable. One important factor is to make sure that
one kind of token does not dominate the coding. This is called determining or defining the variable context, and is
one of the most important decisions we have to make in performing a variation study.
Next, we need to know the best way to measure the linguistic variable. This could be a numeric measurement. For
example, I could use a phonetic analysis program in order to measure the formants of each token, and use the
difference between the two measurements for each formant as the variable. This type of measurement will provide
a fairly fine-grained view of the variation, with theoretically an infinite number of measurements. There
On the other hand, I could simply listen to each token and auditorily decide whether to categorise the variable, so I
would have a categorical or discrete variable. There are advantages to each approach: the interval variable can
presumably capture more subtle levels of variation, and is more objective and not susceptible to uncontrolled biases
of the analyst, as might be the case for the categorical version. On the other hand, it will probably take longer to
perform the coding for the continuous measurement.
We are looking for measures that are reliable and valid. Reliability refers to the extent to which the measures will
be the same if repeated at another time or by another rater. Validity refers to whether the measurement is a ‘true’
or accurate representation of the thing being measured.
One very practical detail that is often asked is exactly how one sets up the record for coding. The easiest way is to
use a spreadsheet program such as Microsoft Excel, a data sheet in a statistics program such as SPSS or SAS, or a
database program such as FileMaker. When setting up a spreadsheet, it is best to have each row represent one
token. Within that row, the first item should be the variant of the linguistic variable, followed by any factors that
you code for, such as linguistic environment or speaker characteristics. Because each row is different token, tokens
from the same speaker will have the same speaker characteristics.
How many factors should you code for? The rule is to code for as much as is practical. Similarly, make as many
distinctions as you can. The three-way distinction for variants of the linguistic variable is similar: I can always collapse
two of these categories so that I only have two variants, but if I had coded only two variants I would need to recode
each token manually to do so. Therefore, we always record for more factors than we think we need, and we always
make as many distinctions or categories as we can.
DESCRIBING PATTERNS
Once we have finished our coding, we count the frequency of one variant for specific factors. In order to address the
constraints, transition, and embedding problems, however, we will want to count in such a way that we can find
patterns. First, we will want to describe the patterns, and then we will want to test whether these patterns are
significantly different from data that have no discernible pattern.
In displaying a simple pattern – that is, a pattern that represents one factor group – tables are the most useful form.
First, everything should be labelled transparently. Because we are interested in relative frequencies, the data are
represented as percentages for each factor. In constructing tables of percentages, one should always show how the
percentage total to 100, and also show the number of tokens each percentage represents.
Tables are useful for relatively straightforward patterns when only one or two factors are involved. However, a
simple graph is often the best way to show patterns when two or more factors are involved.
Finally, we could simplify the chart by viewing the frequency of only one variant. In fact, if a variable has only two
variants (a binary variable) then a graph showing the frequency of one of them is the best way to present our data.
Traditionally, it is the ‘new’ or ‘vernacular’ variable, or the output of the variable rule, that is grasped.
It should be clear by this point that tables and graphs are not dispassionate displays of objective data, but in many
ways create the patterns that you ‘find’ in your data. Therefore, make sure you view your data in multiple ways.
More importantly, when presenting your data in written work, choose the display that contributes best to the
patterns you see in the data.
Finally, when using tables and graphs, do not simply place the graph in the paper and assume the reader will interpret
it in the same way as you. You must explicate the table or graph so that it fits into the patterns you want to point
out in your narrative. You will be helping the reader to understand the patterns in your data so that you can discuss
them.
Now that we have explored some ways of seeing the patterns of percentages in our data, we are going to need to
ask ourselves whether the patterns we see are significant. There are two ways to think about significance.
• The first is statistical significance, which usually means something like ‘This pattern is very unlikely to have
come about by chance – there must be some reason why the linguistic variable patterns this way’.
• The other kind of significance is practical significance. One important factor about statistics is that, given
enough tokens or speakers of a variable, it becomes unlikely that our statistics will tell us that something is
not significant.
Before we begin, we must remember that statistical tests tell us only how the variable patterns, what constraints
there are on it, and how it is embedded in the speech community. Statistical procedures are only as useful as the
user is knowledgeable, not only about statistics but also about linguistic and social theory. Before we run statistics,
we need to have coded for the most plausible factors to have a large effect on the linguistic variable, and once the
statistics have provided us with a numerical model of the variation, we will need to add a qualitative, possibly
theoretical, interpretation to those patterns that the model represents.
There are several ways that statistics can be used to help describe and verify sociolinguistic patterns. One thing we
want to know about our pattern is that if we find a difference between or among groups, is that difference likely to
have been found by chance or not? If we had measured the population – every utterance, or at least utterances of
everyone in our speech community – then we would know that the difference is not by chance, because we
measured everything. To test for this, we need to know how likely it is that the mean of our sample is the mean we
would measure if we measured the entire population. Fortunately, statisticians have discovered that if we have a
certain number of observations (tokens), we can predict how likely our pattern is to be the one of the actual
population.
There are many ways to describe the patterns in data. Two of the most important for sociolinguistics are the mean
and the median. The mean is the average (all tokens added up and then divided by the number of tokens), while the
median is the ‘middle’ of the data (there are as many tokens larger than the median as there are smaller). We often
use these measures to compare groups and describe patterns.
So, we want to know, for example, whether the difference between men and women in the (aʊ) data is one that is
just because we were ‘lucky’ or because there is a real pattern. We do this by determining the range of values that
are the most likely for the ‘real’ mean for each group (called the confidence interval), and then determining how
much those two ranges overlap. If they don’t overlap, or don’t overlap much, then we can be sure that the rates of
the two groups are actually different. To do this we need to know how many tokens we have for each group, because
if we have more tokens, then the likelihood that we were just lucky is lower. The t-test-is one of the statistical tests
that makes these comparisons. This process of comparing between groups is conceptually how most statistical tests
work, although statisticians have developed many different ways of doing the equivalent of making confidence
intervals and comparing them.
Another way of testing data is to look at how they are distributed, and to see if that distribution is different from
what would be expected if everything were left entirely to chance. This is the basic idea behind the chi-square test.
In this test, you have a table of counts such as this.
In order to decide whether this distribution has come about by chance, you work out what distribution would be
expected if the data were distributed by chance and find the difference between the observed and expected
distributions. On the basis of the number of factors you have, you can look up the probability that the difference is
significant (because these kinds of numbers are distributed in a way that statisticians can predict).
A similar procedure for different kinds of data is the Analysis of Variance, or ANOVA. This is like the chi-square test
in that you are trying to decide whether the pattern is not due to chance, but it is used when we have interval data
for the linguistic variable and categories for the factors. It is similar to the t-test above, but it can handle more than
two groups. The test can say whether the different categories are distributed similarly or are different, and then you
need to decide where the difference is.
While these tests are useful possible tools for variationist analysis, they all lack the ability to decide which factors
are having effects on the linguistic variable. This is where the ideal of modelling comes in.
The simple idea is to write an equation that uses the factors we have coded for to predict the rate of our linguistic
variable. We then compare the prediction and the real data. The best such equation will give us the size of the effect
of each factor.
Varbrul (also known as Goldvarb and most recently Rbrul) is the dominant statistical procedure used by variationists.
It works by creating a model, and then comparing it to the real data. It then changes the model slightly to see if the
second is significantly different, and if it is, it keeps this later model and tweaks it. This comparison goes on until the
program cannot do anything to make the model fil the data better. The output is a series of probability weighs that
tell us how important different factors are, and in what direction they change the rate of our linguistic variable.
When Sankoff and Cedergren first developed this program, the statistical technique – logical regression – was quite
new. In the intervening years, the field of statistics has advanced considerable, to the point that it can be argued
that variationists should be using a more modern technique known as mixed effects modelling. Mixed effects
modelling is a technique like multiple regression, but it models factors differently depending on their status as so
called fixed or random effects. A fixed effect is essentially a factor that the experimenter controls, and suspects has
an effect on the linguistic variable. A random effect is one that might have an effect on the variable, but the effect
is not predictable in the way that fixed effects are. In general, in a variation study, our main random effect will be
individual speakers, while our fixed effects will be linguistic factors and social grouping factors such as age, gender,
and class. What the mixed effects model does is factor out the effect of the individual speaker from the effect of the
speaker’s group. all of these sources use the R statistical package, which has the advantages of being free and
computer-platform independent. We sue these statistics to explore which of the patterns that we see in the data
are significant. The meaning of those significant patterns depends on our understanding of how the patterns we
found relate to those that have been found in other variation studies.
GLOSSARY
Apparent-time studies – Those are studies that compare speakers of different ages or groups recorded at a single
time. If there are significant differences in the use of a variable from the older to the younger generation, then it is
hypothesised that we are seeing a change in progress in the direction of the younger’s usage.
Coding variables - Coding the speech means measuring and recording each token of the linguistic variable and
recording any social or linguistic variables that go with that linguistic variable. The first decision is to determine what
counts as the variable. One important factor is to make sure that one kind of token does not dominate the coding.
Community of practice – It is a group of people that is organised around some practice or activity, such as chess or
book club. The practice can be explicitly socially oriented, so that a group of people who dress similarly or have
similar musical tastes can form a community of practice. Such a community is similar to a network, with the
difference that the community of practice is organised around the common practice, rather than around individual
ties to other individuals.
Ethical linguistics - Sociolinguists have been concerned with helping the people who agree to provide data. This
concern begins with the permission to record someone's speech. They have long argued that our ethics can go
further. Speakers need to know more than the simple fact that they are being recorded. Some argue that in fact we
should have a net positive effect on the communities we study.
Judgement sample - This kind of sample is as random as possible but also determines ahead of time how many
people of certain characteristics should be included in the sample, often on the basis of the proportion of that
characteristic in the population as a whole.
Panel study (same speakers) - They are real-studies of variation across real time when the participants are held
constant and where the same people is interviewed again after a period of time. By doing that, the researchers can
learn about the possible change of a given speech community, while trend studies involve comparing speech from
members of the same community at different points in time, but they are not the same respondents.
Principle of error correction - It states that "a scientist who becomes aware of a wide spread idea or social practice
with important consequences that is invalidated by his/her own data is obligated to bring this error to the attention
of the widest possible audience".
Principle of debt incurred - It states that "an investigator who has obtained linguistic data from members of a speech
community has an obligation to use the knowledge based on that data for the benefit of the community, when it
has need of it".
Principle of linguistic gratuity - It states that "investigators who have obtained linguistic data from members of a
speech community should actively pursue positive ways in which they can return linguistic favours to the
community".
Random sample - It is the most statistically rigorous kind of sample, in which a certain number of people are chosen
from a population (speech community) completely at random, perhaps by choosing telephone numbers randomly
in a particular location.
Real-time studies - Studies that compare speech recorded at different times. They give us the most information
about changes in a speech community.
Reliability - It refers to the extent to which the measures will be the same if repeated at another time by another
rater.
Sample - Since we cannot gather data from everyone in a population, we have to gather it from a small group of
such population, the so-called sample, which is as close as possible to the population in every characteristic.
Shared linguistic norm - These shared norms can refer to similar ways of producing speech, but also often refer to
language more generally. The shared norms thus do not entail everyone speaking alike, but could mean that
speakers produce language differently and at the same time evaluate speech in a similar manner.
Social networks - They are groups of people who have some sort of social tie to one another, and in that sense a
network could be seen as a kind of speech community. A network sampling is one method often used by
sociolinguists. They start with a few individuals whom they already have met in the community or introduce
themselves, and then ask them to introduce them to friends and keep asking for introductions.
Trend study (different speakers) - It involves comparing speech from members of the same community at different
points in time.
Validity - It refers to whether the measurement is a "true" or accurate representation of the thing being measured.
Social factors lie mostly in the realm of the embedding and evaluation questions: ‘How are variation and change
embedded in the social fabric of the speech community?’ and ‘How does the speech community evaluate the
variables and the change?’. The work of describing the embedding of variations is the central descriptive work of
research on social factors. We will mainly focus on ‘canonical patterns’ and challenges to them. The focus on the
embedding question dovetails with the cognitive factors involved in variation, and how those uses, and meanings
arise.
Labov approaches the problem of social constraints by using sociolinguistic methods to track down the ‘leaders’ of
sound changes – those speakers who use the most of a new variant at any one time, and whom the rest of the
community appear to ‘follow’ in later generations. The embedding and evaluation problems are bout more than just
the leaders, because the entire organisation of variation is embedded and evaluated in the speech community. If it
were as simple as everyone following the leaders, the patterns would not be so difficult to find and organise.
However, we will see that the patterns are not ‘laws’ of sound change, and even the principles can be fairly weak.
We thus need to understand the embedding of variation in the community from the entire community’s perspective,
rather than just of the ‘popular kids’.
We thus need to understand what the social motivations for the linguistic decisions that speakers make are, and
what are the processes that make that meaning.
When variationists have discussed social factors, they have traditionally referred to patterns that predict rates of
variant usage for major ‘identity categories’ such as age, class, sex/gender, race/ethnicity, and religion. This view of
the social space is generally referred to as structuralist in sociological theory, a usage which derives from the
linguistics one, in that both refer to a system of oppositions, in which the identity of an item is dependent upon and
entails not being something else. Structuralism in sociology refers to a view of society as having a structure of
distinctions in a system. So being a man means not being a woman, and being upper-class means not being working-
class; the terms are only meaningful in relation to their structural relationship to the other terms or positions in
society.
Despite the criticisms, we will see that in study after study researchers have shown that one can discern patterns
such that certain structural social identities in a speech community will correlate with the use of one variant over
another, and that some of these patterns goes so far as to claim that ‘correlating linguistic variation as the dependent
variable with independent variables such as linguistic environment, style or social categories is the primary task of
sociolinguistics’. It is true that correlations are the main kinds of patterns that variation researchers attempt to
discover in the pursuit of answering the embedding question. Depending on how it is framed, this question could
even be answered by a listing of patterns of change based on these structural social identity categories.
While we might be able to find some consistent social patterns such that one of these groups tends always to pattern
in the same way, these patterns are organised into three processes:
• Stratification is the process in which language is used differently by groups who are placed on an ordered
prestige scale. Socioeconomic class is the prototype of stratification. However, the term refers to cases
where stratification and value are inherent in the categories, such as caste and class. Stratification also
inherently implies a scale in which certain kinds of speech are valued more highly than others.
• Accommodation is the process whereby we tend to speak more like the person we are speaking to.
• Differentiation is in a sense the opposite process to accommodation, whereby a group of speakers change
their variable usage so that it is different from another group’s.
The difference between stratification and differentiation is that in differentiation there is not necessarily an implied
consensus in the community about the value of speech for one group or another.
STRATIFICATION
Labov’s pioneering work focuses the variationist enterprise on stratification. Labov’s study relied on a number of
interlocking studies showing that language in New York City was stratified both in terms of production and in terms
of the evaluative norms that speakers had for such speech.
The best-known of these studies is probably Labov’s of (r) as used by employees of department stores. These stores
were used as indirect measures of class, with the assumption that the salespeople in the stores were more likely to
present themselves linguistically as desirable to the different clienteles. Labov identified three department stores,
determining later their class stratification by comparing prices in advertisements and the readers of papers they
advertised in, and also through more qualitative measures such as the appearance of the stores. Labov found that
the stratification regularly patterned such that there was more ‘r-lessness’ in the lower-class stores, and there was
the most ‘r-fulness in the upper-class store. Labov thus found evidence that r-lessness is higher for the working class
than for the upper-class. He also found a pattern by age. Taking these together, Labov argued that in New York City
the chance was beginning in the upper class and then moving to the middle and working classes.
The other major part of the survey was based on a sample of the population of a federally founded programme
addressing ‘juvenile delinquency’. The class measures were based on a ten-point scale used by the survey and serves
as a good example of how class stratification is often measured. The scale is based on three indicators, each of which
contributes to the number of points a person receives: occupational rank, educational rank and income rank.
Education was measured on levels completed, while income was measured on relationship to the nation’s median
yearly income. Occupations were divided roughly into professionals, clerks and salesmen, skilled workers and blue-
collar workers, and finally service workers and labourers.
Labov’s hypothesis was that the rate of variation would correlate with class in a linear way. He did not expect, for
example, that the lowest class would pattern with the upper class and the middle classes together; hence the term
‘stratification’. Labov indeed found clear stratification patterns for each variable.
In order to compare speech among respondents, Labov noted famously that it is not always possible to distinguish
‘a casual salesman from a careful pipefitter’. So, we need to control for the casualness or carefulness of speech;
basically, controlling in some way for intraspeaker variation. Labov addressed this problem by assuming that more
casual speech is speech in which the speaker is not paying attention to the speech itself. So, the most careful speech
will be when uttering a minimal pair highlighting the variable in question, because it draws attention to the variable.
The most casual speech is when telling personal narratives, because attention is focused on the action in the story.
Labov found that, in general, social class stratification if style is controlled for, and within each social class, styles are
ordered in the same way.
In addition, each class uses more re-fulness as the style becomes more careful. However, in the most careful styles,
the lower middle class actually uses more (r) than the upper middle class, a finding which has since been termed the
lower-middle class hypercorrection pattern.
We also wish to know whether change is occurring and if so in what direction. Real-time sources suggest that before
the middle of the 20th century, New York was mostly r-less. Because this new form is used more by more upper-class
speakers, we say that the upper middle class is leading the change. This type of change has been dubbed a change
from above, because it seems to begin in the upper social classes, and also (it is claimed) is a variable about which
speakers are aware (it is above the level of awareness).
Similar patterns were found by Peter Trudgill in his survey of Norwich a few years later. Trudgill’s measure of class
was similar to Labov’s, and he found similar patterns of stratification. However, rather than the fairly even, linear
stratification that Labov found, Trudgill found small differences within his middle-class categories, but between the
lower middle classes, he found a stark separation.
Guy suggests that these differences may not be random but related to the status of the variables in terms of
awareness and meaning. This observation raises an important distinction made by Labov early on between change
from above and change from below. These terms refer to ‘above’ and ‘below’ in two ways. First, they describe the
so-called awareness of variables, in which variables that the community is aware of are changes from above, and
variables that the community is not aware of are changes from below. Changes from above are also thought of as
changes that tend to be led by the upper classes, and changes from below are led by lower classes. However, they
are extremely important concepts in variationist research, because different patterns and motivations for these
different changes have been found. More recently, Labov has argued that change from above is a change by
diffusion, which is characterised by a spread of one variant from one geographic area to another, mostly by adults,
while a change from below is change by transmission, in which a change is under way and is moved forwards by
children as they acquire the variety.
Horvath also finds class differences in her study of the vowel system of the Sydney speech community. She identified
four ‘sociolects’ based on four groups of speakers. She then looked at the composition of the sociolects, and found
difference for gender, age, ethnicity, and especially class. Her work is important because it shows that even if we
code all speakers just by their speech and not by class ahead of time, we will find a class stratification pattern.
One question we might ask is whether any components of the class measures are more important than the others.
Labov attempts to determine which components best predict the sociolinguistic variation, and he finds that of the
individual indicators, occupation is often, but not always, the most robustly correlated factor of the components of
the socioeconomic index. He finds that occupation is the single measure that most reliably stratifies speakers, as
opposed to measures of achievement or quality such as income or education. This is an important finding, because
it suggests that occupation is reflecting something about people, in that it is more closely related to how they talk
than income or education. The fact that occupation is the stratificational measure makes sense if we notice that it is
the measure that is most about what people do and how they relate to people, rather than what they are or have
done.
All of the examples given above are phonological variables. Do similar terms hold for morphological or syntactic
variables? There is evidence that they do. Multiple negation, also referred to as negative concord, in English is one
of the most sensitive variable to social class. Labov’s study of negative concord in Philadelphia shows the sensitivity
of this variable to class. The lower working classes uses much more negative concord than any other class. Wolfram
found a similar pattern for African American speakers in Detroit.
It is clear from these studies that stratification is a central finding in variationist work. Where there is a ‘vernacular’
or ‘non-stander’ variant, that variant tends to be used more by speakers classified at the lower end of the class or
prestige structure. However, we have not explored patterns of interaction. Interactions are patterns in which two
categories do not act independently. For example, we might find that upper-class men do something very different
from upper-class women, and that this upper-class gender difference does not hold for the working class.
Even without interactions, it is not entirely predictable from class how a change will be embedded in a speech
community. Labov argues that generalisations can be made if we divide changes in progress into our two types:
change from below and change from above. In change from below, the leading class will be a low but ‘interior’ class.
That is, the leaders will be working-class speakers but not the lowest-class speakers. However, a change from above
will be led by the middle or upper middle class. Labov proposes to capture this generalisation in the Curvilinear
Principle: ‘Linguistic changes from below originate in a central social group, located in the interior of the
socioeconomic hierarchy’.
Another way that his stratification has been viewed is through the notion of the linguistic marketplace. Labov argues
that prestige is attached to different linguistic forms. The ways of speaking can represent different amounts of what
Bourdieu calls ‘symbolic capital’ in a speech community. We see then that the patterns of class stratification can
represent different values of persons in some market – often the labour market. In fact, all of these stratification
patterns suggest a representation of some kind of social meaning for different variants. However, the correlations
do not provide the meanings.
Another way of thinking about stratification is geographically. First, it has been shown that changes generally tend
to originate in populous urban areas and spread from there. There are two models of how this diffusion might work:
either gradually over geographic space (the wave model), or from the most populous cities and then to the second-
most-populous cities, skipping over very small towns, which adopt the change last (the gravity model). The wave
model is less common than the gravity model. Both of these models indicate a pattern in which the urban centre,
which often has the most prestige and cultural capital, is in a way equivalent to the upper-class speakers. These
kinds of changes are usually represented as mechanical effects of contact rather than involving issues such as
prestige. That is, because of the kinds of communication and contact that medium-sized towns have with larger
ones, there is a ‘natural’ spread (perhaps by accommodation) from one to the other. This view would see this as a
case of accommodation. However, we can also see it as a possible stratification, where people are identifying with
the values, lifestyle, practices, or prestige of the larger urban centre, and not simple automatically accommodating
to it.
There have also been diffusion patterns that do not fit these models, however. Bailey found one variable that was
moving from smaller towns to larger urban areas. In this case, however, Bailey showed that the term is associated
with a renewed prestige attached to the traditional southern identity of the Texas English. So, our view that diffusion
is a kind of stratification is upheld.
Horvath and Horvath, however, present more problematic data. Their work suggests that in a more global
perspective on diffusion, these patterns do not seem to hold. Horvath and Horvath analyse /l/-vocalisation in
Australia and New Zealand. Either the wave model or the gravity model would predict that a change in these
countries would proceed from the largest cities to the smaller, and from the larger country to the smaller. Horvath
and Horvath find that neither is the case. In fact, New Zealand is the leader as far as /l/-vocalisation goes, and the
smaller cities in both countries use more vocalisation.
There is a second manner in which geography can be considered to be stratification. The patterns we characterised
as change from above and change from below are often reflective of geographical and standardisation stratification
patterns. Change from above tends to involve changes that have as their source a less ‘localised’ variety, perhaps
even a national standard. The stratification of (r) is a good example of this pattern. R-lessness in New York, and the
US more generally, was a feature of upper-class speech until after World War II. After the War, a new national
‘standard’ form in r-fulness developed, and it is this new, more national way of speaking that then made its way into
New York.
In a sense, then, there is an element of variety contact happening in such a case such as that of (r) in New York: a
local variety and a less local variety. Globalisation, media, and mobility are playing an important role in the
development of varieties in the developed world, to an extent probably not occurring previously. This can be seen
in Pittsburgh, Pennsylvania, where the local variant is clearly being used less and less by younger generations.
However, this loss has come about just as Pittsburghers have begun to talk about the local variety as a known
‘language’ that they refer to as Pittsburghese. But even as the local variety becomes more and more associated with
a local identity, the everyday use of such variables is falling away. Dubois and Horvath argue that changes from
above and below should be viewed as exonorms and endonorms, respectively, because so-called changes from above
are most often norms that are adopted from contact with non-local varieties.
ACCOMMODATION
Language can be used to create solidarity relationships in a speech community. In fact, these solidarity relationships
may in many ways be stronger than the stratification patterns or even the differentiation patterns.
• First, accommodation takes place among individuals in the short term, in conversations, when speakers
tend to speak more like the person they are speaking to.
• Second, we find that over longer stretches of time, people make more enduring adjustments to speak in
ways similar to other people whom they regard as close friends, otherwise known as their social networks.
• Finally, when large groups of people come into contact, such as in case of large-scale migration, we find
that in general the members of the groups end up speaking in ways that are similar to each other and
eventually form new varieties known as koines.
Lesley and Milroy were the first to use network analyses in a large-scale sociolinguistic survey. They showed that we
can correlate speakers’ variable use depending on the kinds of social networks that they are in. The researchers
characterised networks in two dimensions. First, networks can be measured in terms of density, such that dense
networks are those in which people have connections, or ties, to many other people in the network.
Density can be calculated numerically in a number of ways, but the most common is a network strength scale of
some sort, which is a composite of a number of measures such as the number of ties. Within a network, we can also
measure the different kinds of ties that people have to each other. The number of different kinds of ties is a measure
of the multiplexity of the network, such that multiplex ties or networks are those that have many such connections.
Milroy showed that the kinds of networks that speakers are in can predict variable usage, such that dense and
multiplex networks tend to be correlated with the use of a local, or vernacular variant.
Cheshire measured the integration of young boys and girls into a peer group network and found that those who
were more integrated used more local morphosyntactic variants. She determined that the ‘core’ members of an
adventure-playground group used more ‘vernacular’ variants than those who were ‘secondary’ of ‘peripheral’.
Network status was determined from two criteria. The first was obtained by asking the kids whom they spent the
most time with. Second, Cheshire devised a ‘vernacular culture index’, which was based on social practices and
aspirations, crime participation, etc. The core members all showed the most vernacular practices and were also
named the most frequently by others in the network. Cheshire thus began to show what networks strengths
measure – the strength with which a person is socially integrated into the local practices of their community.
It is important to note that network strength is not universally predictive of vernacular language. For example, Salami
studied the effect of networks on the use of six phonological variables in the variety of Yoruba spoken in Nigeria. For
two variables, networks strength had an effect. However, when Salami tested whether network strength had an
effect separately for men and women, he found that the networks correlated only with the variables for women.
Thus, network does not have a consistent effect across all groups in every study. Network patterns are thus useful
and often predictive of variation, but this effect can often be very complex, and is not a universal tendency. Network
patterns are thus useful and often predictive of variation, but this effect can often be very complex, and is not a
universal tendency.
Network analyses point to the complex relationship between accommodation and stratification in variation studies.
This relationship is highlighted when we ask whose language is accommodated to. That is, are there some speakers
in a network wo are the real innovators, and others in the network who are the accommodators? Labov shows that
there are central actors, or ‘stars’, who have the greatest influence on their network. Eckert shows that certain
individuals become, as she calls them, ‘sociolinguistic icons’. These icons are the reference points to which other
students accommodate. If stratification represents differences in the prestige attached to variants, then these icons
have a kind of prestige, and represent a more local way that stratification is replicated. More importantly, it is not
only membership in a dense or multiplex network, but a particular position in the network, that is important in
predicting an individual’s use of a variant.
One of the best general pieces of evidence that accommodation is an important force in linguistic change comes
from the studies of koines, which are generally discussed as new dialects that form from a mixture of previous
varieties. Bortoni-Ricardo investigated the process of urbanisation in Brazlandia, Brazil, and the influences on
speakers adopting features of the most vernacular Brazilian Portuguese variety, known as Caipira. Bortoni-Ricardo
was interested in measuring the difference in how insulated from or integrated with the urban milieu the rural
migrants were. For this study, Bortoni-Ricardo had two network measures. The integration index was a complex
measurement of network strength, while the urbanisation index was a measure of the urbanisation of a speaker’s
network. Bortoni-Ricardo found that although the urbanisation index had predictive effect for both men and women,
the integration measure was different for men and women. Investigating further, Bortoni-Ricardo found that for
men, the two measures were positively correlated with each other, meaning essentially that the indexes were
measuring similar social aspects, while for women, the indexes were negatively correlated, meaning they were
opposite aspects. Bortoni-Ricardo explains this result with the observation that a strong integration measure for
women is more likely to be due to dense networks composed of other rural migrants (and family members), while
when men have strong networks, these are not necessarily made up of other migrants. So, we again see that
accommodation, through a measure of networks, operates to affect linguistic choice, although other measures
would be needed to operationalise the kinds of networks involved.
Kerswill and Williams also provide convincing evidence of this network effect in another study of migration and
dialect contact. This study focuses on the development of a koine in the ‘New Town’ of Milton Keynes. This study
has been extremely important for informing our knowledge about dialect contact and variation in general, especially
since the design included young children around age four and older children around age twelve. Kerswill and
William’s qualitative look at several of the older children who were in the forefront of the new dialect is revealing.
The researchers concluded that ‘the main factor is the child’s orientation towards the peer group. All the high scorers
are very well integrated into a group of friends; they are sociable and often cited as friends by other children. By
contrast, the low scorers are somewhat distanced from their peers’. Kerswill and Williams also confirmed that this
network effect was far stronger than any connection between the speech of the caregivers of the children and their
scorers. These results suggest that it is the children with more dense and multiplex networks who are more involved
in creating the ‘new’ variety in Milton Keynes.
So, it is pretty clear that we can predict who will be the movers and shakers in variation: those people who are more
central to social networks. Moreover, subgroups or speech communities with more dense and multiplex networks
will be more likely to be affected by the norms in the group than those who have looser networks. We might ask,
then, whether we might be able to explain the class differences we find through network differences – perhaps
working-class speakers are more likely than upper-class speakers to have more dense and multiplex networks – or,
whether any pattern we discuss can be explained in terms of networks. Milroy and Milroy cite that ‘Generally
speaking, middle-class networks (consisting largely of weak ties) are larger, less kin – and territory – oriented and
perceived as more supportive. Mewett argues that class differences in small communities begin to emerge over time
as the proportion of multiplex relationships declines’.
Labov, however, compares regression analyses for ongoing changes in Philadelphia vowels both with and without a
network measure. He finds that while there may be some correlation between network strength and class, it looks
as if they are accounting for linguistic variation in different ways.
Labov and Milroy and Milroy both find support for another finding of the Milroys in Belfast, however, in the role of
what they call weak ties in the diffusion of linguistic changes. Weak ties are the opposite of the strong ties found in
dense multiplex networks. These weak ties are ties of acquaintance – for example, the parent of a schoolmate of
one’s child seen three times a year. Milroys argue ‘not only that groups linked internally mainly by relatively weak
ties are susceptible to innovation, but also that innovations between groups are generally transmitted by means of
weak rather than strong network ties’. Labov finds that the leaders of linguistic change in Philadelphia are
characterised by being central in their own networks and are ‘not limited to their local networks but have intimate
friends in the wider neighbourhood’.
In sum, let’s review the ways that social networks have been operationalised in the sociolinguistic literature. these
have all been productive in predicting who will use how much of a particular variant. These measures include:
• Density (the number of ties) and multiplexity (the number of different ties linking two people, or an
aggregate of that for a network). Both correlate with vernacular variants. Group members accommodate
to each other more readily if they all know each other in multiple ways.
• Weak and strong ties. While dense and multiples networks reflect strong ties that correlate with vernacular
use, weak ties are theorised to promote the diffusion of a linguistic change. Group members are more likely
to accommodate to people who have status in the group.
• Identity of the network. Finally, there is evidence that the kinds of ties are not the only predictor. Bortoni-
Ricardo’s show us that the norms that are accommodated to within a network can depend on who is in that
network.
In general, then, it is fairly clear that accommodation is one of the basic processes of variation: people tend to talk
like the people they are talking to. They talk more like people they are predisposed to like, and people they are
otherwise close to socially, but they also like to ‘borrow’ from people they don’t know well.
DIFFERENTIATION
While we can find differentiation in many situations involving variation, there are three main social domains in which
differentiation has been found: geography, race and ethnicity, and gender. In each of these domains, we find that
some difference in identity or place leads to speakers making different linguistic choices from other speakers. We
might assume that differentiation is just the flip-side of accommodation – since we cannot accommodate to
everyone, we must end up with some group differences.
In the case of geographical differentiation, when there is a difference in language or variety across boundaries such
as mountains or rivers, it is pretty clear that there are fewer opportunities for accommodation across these kinds of
boundaries. But there are some mysteries that suggest the answers is not so easy, such as why geographical
boundaries seem to be stable for generations, even without clear physical boundaries.
Britain’s study of the English Fens provides us with a view into how such boundaries develop, change, and are
maintained. The Fens are an are of southern England whose land was a march until a few hundred years ago, when
it started to be drained and become productive farmland. One can understand why, over the thousand of years
preceding this, differences developed in the English spoken to the east and west of the Fens. Nevertheless, a
boundary has been maintained up to the present day. Britain shows that there is very little interaction between
people on either side of the dialect boundary, whilst few people live in the boundary area, and transportation
networks favour local and north-south routers rather than east-west connections between the two major towns of
the area. The links between the populations in the east and west are therefore weak, and this is one factor leading
to their differentiation. However, we might expect that in fact there would be some accommodation between the
two areas. The difference is that there are also ‘local rivalries and negative stereotyping of each other’s residents’.
So, there is what we can loosely call an ideological or identity differentiation that reinforces the physical and social
practice differentiation.
Some boundaries, though, are still somewhat mysterious. In North America, there are vast expanses of the continent
that essentially share a variety, separated by thousands of miles. For example, Chicago, Detroit, Cleveland, Buffalo,
and Rochester all share a dialect region. Pittsburgh and Columbus, however, which are both closer to Cleveland than
Cleveland is to Chicago, are in quite different dialect regions, with little in the way of rivers and mountains in
between. It thus appears that differentiation is not something that is automatic on the basis of geography, but
incorporates a dimension of social practice, and an ideological or identity dimension.
A view of race or ethnicity presents a similar and starker picture of ideological and identity differentiation: we find
people living near each other and often working with each other, but nevertheless speaking quite differently. The
case of AAVE is the most striking example of this. AAVE is a case of both uniformity and differentiation. It shows
uniformity in that it is a variety that is very similar across many cities in the USA. This similarity points to some of
supraregional norm or process – that allows this to be described as a single variety in most cases. The remarkable
fact about AAVE is that it shows so much difference from non-Black varieties spoken in the same cities and towns.
Wolfram summarises a number of studies of communities that show older speakers using more local Anglo-
American features and younger African American speakers increasing in the use of some linguistic variants
associated with AAVE. He notes that some these changes may be due to expanded mobility for younger speakers,
so they are exposed to AAVE whereas the older speakers were not. However, expanded mobility by itself does not
produce this difference – the speakers still must orient themselves to these new variants, and make choices to adopt
them in their own speech.
Of course, AAVE is not the only example of such ethnic differentiation, and immigration is not always the only cause.
The other two main ways that ethnic groups come into contact are through voluntary migration and through
colonialism. The former is the more common and happens quite often. In these cases, there is an added complication
that aspects of the ancestral language may have an ‘interference’ or ‘transference’ effect on the language. This
language contact phenomenon, especially among second-generation members of the ethnic community has been
documented to some degree, although it is often difficult to show without any doubt that a second-generation
speaker uses particular variants because they are bilingual. That is, it is almost impossible to show that someone
who speaks a language ‘defectively’ does so because of their bilingualism. Indians have long migrated to South Africa,
and those who migrated were generally not native English speakers. As English is spoken more and more by Indian
South Africans, a variety seems to be developing, and many of the features that it exhibits are clearly ‘transference’
features from Indian languages.
One of the most complex situations when it comes to differentiation and language is that of the migration of Spanish-
speaking groups to the United States. It is important to understand that Spanish has been in the United States about
as long as English. However, there has always been a steady flow of migrants to the USA. Variation in the USA is
complex, due to English-Spanish contact as well as contact among many different Spanish varieties. There is thus
the possibility of an English variety arising, and also the possibility of a koine of several varieties of Spanish.
We might predict that, just as Indians in South Africa have created something of a common language of identity, so
Latinos in the USA would create a kind of koine variety of Spanish as they began to identify as Latinos and not
specifically Mexicans or Puerto Ricans. Indeed, there is some evidence that something like this may be happening in
New York. On the other hand, in Chicago, there is absolutely no accommodation going on because there is an
antagonistic relationship between Puerto Ricans and Mexicans. In fact, there is differentiation both in the Spanish
spoken by the groups, and in whether Spanish is spoken at all.
All of these examples show that ideologies about groups an identity differentiation are important ways that language
variation patterns socially.
The effects of ideology and identity are even clearer when it comes to gender variation. Gender is at once the most
obvious and the most puzzling of social factors affecting variation, and it has come effect almost universally. Gender
differentiation is probably the most recurrent social pattern in sociolinguistics. That is, if a speech community shows
any social differentiation at all, then there is likely to be some gender differentiation. Along with class stratification
and accommodation in networks, gender differentiation creates the main patterns that a theory of variation and
change has to be able to explain.
Labov identified some principles for gender differentiation. Labov’s first formulation of these principles comes from
his article focusing on gender patterns, in which he catalogues a large number of changes to support the principles.
He makes a distinction again between stable variables and those involved in change, and between changes from
above and changes from below, for those undergoing change:
Principle I. For stable sociolinguistic variables, men use a higher frequency of ‘non-standard’ forms than
women do.
Principle Ia. In change from above, women favour the incoming prestige form more than men do.
Principle II. In change from below, women are most often the innovators.
This leads to what he calls the gender paradox, in that women lead in both changes from above and changes from
below.
The greater use of vernacular variants by men was found in one of the first studies of variation, that of the (ing)
variable by Fischer. His study of school children in New England found that boys used more the [ɪn] variant than the
girls. Negative concord (NC) is another variable that often shows this pattern of gender distinction. Eckert found that
men use more NC than women.
Principle Ia states that women lead changes from above. In Pittsburgh, /aʊ/-monophthongisation is clearly used
more by men than women. Thus, we can say that women are leading in this change from above.
Principle II states that women lead in change from below. A good example is Labov’s regression analyses of the vowel
changes in Philadelphia. Out of fourteen vowel variables undergoing change, eleven show that women are in
advance.
These patters are described by Labov as the gender paradox: ‘Women conform more closely than men to
sociolinguistic norms that are overtly prescribed, but conform less than men when they are not’. It is worth
unpacking this statement somewhat, and the assumptions contained in it. First, the statement is formed in terms of
what women do, which gives an impression that men are the norm against which women should be measured, and
that men are acting statically, and women are acting in reaction. However, there’s no reason this paradox could not
be formulated in terms of men. In fact, this behaviour seems more paradoxical, because we would think that people
who don’t follow overt norms might require more explanation than those who do. Next, note that the language of
change from above and below is represented in terms of norms that are ‘overtly prescribed’. However, Labov never
explains what it means to be overtly prescribed, and the exact mechanism of overt prescription. This topic will
formulate much of our discussion in the challenges to the canonical patterns. Finally, we will need to pick apart what
it means to conform: is this purely a quantification issue, or should we be more sensitive to how a variable is used in
particular speech situations?
The patterns discussed in this chapter thus are far those that one finds in most accounts of sociolinguistic variation.
I call them canonical because in the academic discourse of variation and change, these studies and patterns serve as
reference points for newer studies.
We have seen that there are three main overall social patterns of variation: stratification, accommodation, and
differentiation. Another way of thinking about these patterns is the kinds of motivations that have been suggested
for each pattern: prestige, solidarity, and difference. So variationists have argued that in patters of stratification,
some kind of social prestige is attached to one variant and some kind of stigma to another. In accommodation
patterns, speakers are motivated not by the prestige or stigma of the particular variant but by solidarity with other
speakers in their networks. Finally, patterns of differentiation are theorised to arise because speakers want to
contrast their identity with a certain group. The difficulty with these divisions is that at any time, a variant might be
used for any of these purposes, or have any of these kinds of meanings.
There are four ways that the canonical patterns have been complicated and challenged:
• In many studies researchers have found complex interactions among social variables. For example, we
might find that men and women show different rates of a variant in one class, but no difference in another.
• New studies are sometimes performed that provide counter-examples to the canonical patterns.
• We find studies that rethink or refine the analytical terms used (such as class, age, ethnicity, and gender).
• We have to acknowledge studies in which there is no pattern where we would expect it.
It would be convenient if research fell neatly into these categories, but in fact many studies challenge canonical
patterns in more than one way.
The first of these challenges some assumptions made about the concept of class. Rickford’s analysis of class in the
Guyanese village of Cane Walk leads to a questioning of how class is thought of in variation studies. The studies in
the canonical tradition take a view of class that is a consensus, functionalist view. This understanding of class posits
that variants used by the upper classes are valued by both the upper and the lower classes; even though the latter
do not use these variants as much, the consensus view assumes that the lower classes concur with the view of their
value. It should be noted that Labov argues that this is the case on the basis of the hypercorrection pattern and on
‘subjective reaction tests’. The understanding is functionalist in that it views each class stratum as having a functional
role in the workings of society. By contrast, a conflict view argues that people from different classes have different
economic interests in society, and these are often in conflict. Such a view suggests that in fact working-class speakers
may want to resist the prestige that is symbolised by the variants used by the upper-class speakers because the
working classes have no stake in that prestige system.
We do need to question whether all uses of a variable are going to be evaluated on the class continuum in the same
way as they are in the somewhat artificial situations in which speech is gathered for most sociolinguistic studies.
There is a third model: that of hegemony. This captures both the conflict and the consensus models, arguing that
the working classes do value the speech of the upper class, because they have been conditioned to believe that what
is in the interests of the upper classes is in their interest, whereas in actuality is not.
One of Rickford’s main points, however, was that the consensus models are culturally constructed and variable, and
that views of class and the canonical patterns surrounding those views should not be seen as universal, and specially
not as timeless.
Others have shown that not all variables are differentiated in situations where we might predict it. The identity-and-
ideological dimension of race differentiation is most clearly shown in a recent study by Eberhardt in Pittsburgh, in
which two different variables behave differently depending on how much each has ideological associations with
race. Eberhardt was interested in whether the African American community in Pittsburgh was adopting any features
of the White variety or whether they were orienting more towards AAVE. She chose two phonological variables to
investigate, both present in the local White variety but absent in all descriptions of AAVE. One was /aʊ/-
monophthongisation, while the second was the so-called low-back merger (LBM).
Pittsburgh, like most US cities, shows patterns of housing segregation by race. In addition, unlike other cities in which
there is one are that is identified as an African American neighbourhood, Pittsburgh has several such
neighbourhoods, which are also somewhat distinguished by class. The history led Eberhardt to suspect that there
might have been contact and interaction to lead Pittsburgh African Americans to use some linguistic features of the
local variety. The analysis of the LBM supported this hypothesis: all but one of the African American speakers that
Eberhardt analysed showed the merger, both in production and in perception. This result suggests that in fact the
two group’s varieties are converging, and that there is less difference in Black and White varieties in Pittsburgh than
elsewhere. However, Eberhardt found that no African American speakers showed evidence of monophthongisation,
suggesting the opposite conclusion of that for the LBM.
The key to the difference is the status of these two variables in the community. Monophthongisation is a change
from above the level of consciousness, while the LBM is not. In fact, monophthongisation is the subject of lots of
talk by White Pittsburghers. On the other hand, ‘the low-back merger is not the subject of metalinguistic talk in the
city. The distinction of the two variables show quite neatly that the resistance of African American speakers to
monophthongisation is not a mechanical effect such as limited contact or networks. The difference is due to identity
and ideological factors.
Penelope Eckert shows a rethinking of how interactions among class and gender undermine the canonical patterns.
In fact, Eckert argue that the notion that women’s speech is more standard is a ‘hall of mirrors’. Eckert found that in
a number of ways a high school resolved into two social poles – jocks and burnouts. The jocks were students who
were aligned with the corporate, college-bound culture of the school, and participated more in school activities,
while the burnouts were primarily oriented towards concerns outside school. Eckert shows that a number of social
practices and norms define the poles of these groups. She argues that linguistic variation is a social practice like
others, used to create personal styles in the same way as dress and activities. Note that while there is some
correlation between ‘jock’ and ‘middle-class’ and between ‘burnout’ and ‘working-class’, Eckert finds that class
measures are not as robust as the locally relevant jock and burnout categories. Eckert analysed the use of negative
concord. Here we see that there is a big difference for gender in the jock category, but almost no difference in the
burnout category.
Labov finds similar patterns for some variables. For example, for (ing), he finds a big gender gap in the middle classes,
but almost no difference in the working classes. The pattern is even more striking in the ‘change from below’, which
shows in the lower occupational groups, women use more raised variants, while in the upper occupations, this
pattern is reserved, and men use more.
So, for the changes from below, the generalisation we can make, if any, is that women tend to be at the extremes
of the status spectrum, and it is this more general pattern that needs to be explained rather than the conformity or
standardness of women.
Eckert also differentiates variables in terms not of whether they are changing or not, but of what stage of change
they are in. Eckert shows that the differences in gender appear only in the older, while the newer changes are
differentiated along jock-burnout lines. Here again, our simple categories of change are complicated in ways that
lead to a more sensitive understanding of the embedding of variation in the speech community.
Two more studies challenge how we think about stratification and the relationship of gender to it. Sidnell performed
ethnographic research in a village in Guyana, and analysed the gendered use of the same variable. He found that
men tended to use the more standard ai variant more than women did. He argued that this difference had to do
with the fact that when women used ai too much they ran the risk of seeming ‘uppity’, and so they avoided it. Haeri
found men using more of a standard variant as well, but for different reasons.
Another innovation in Eckert’s work is how she groups her speakers. Her work introduces the concept of the
community of practice. Jocks and burnouts form different communities of practice, because they engage in different
practices and have different norms in many social domains. Eckert explains the concept:
A Community of Practice is an aggregate of people who come together around some enterprise. United by this
common enterprise, people come to develop and share ways of doing things, ways of talking, beliefs, values as a
function of their joint engagement in activity. Simultaneously, social relationships form around the activities and
activities form around relationships.
This concept provides a level of organisation for the creation of norms that is similar to social networks, although it
is focused on the group rather than in the individuals in the group, and the meaning and practices represented by
social networks ties. Eckert’s methods have inspired a number of similar studies. Cheshire found that the degree of
engagement in ‘vernacular culture’ was an important predictor for a number of variables. Her vernacular culture is
a good example of a community of practice, as the kinds oriented in this way engaged in practices stereotypically
associated with the working class.
Dubois and Horvath offer some important critiques of and innovations to almost all of the canon: they question the
notion of change from above and below, and even the idea of a stable variable, and of course the gender pattern
outlined by Labov. They show the importance of a deep understanding of the community being analysed, but they
show how the local history and culture can produce a complex interaction of social factors that suggests
generalisations about these social factors are premature.
Dubois and Horvath also argue that we need to use more precise terms than ‘change from above’ and ‘change from
below’. These researchers identify changes as either adoption or innovation. Adoption occurs when the incoming
variant is a norm from another speech community. Innovation, on the other hand, concerns changes from within
the speech community. This language is similar to Labov’s distinction between change by transmission and by
diffusion. Dubois and Horvath use this distinction more productively, however, to line up historical events with the
adoption and recycling of the variants, where recycling refers to the reuse of variants that were on the wane.
We might be tempted to generalise, then, that men are more likely to recycle older innovations, but there is probably
not enough evidence to make such a broad claim. As Dubois and Horvath show, we need to take into account the
local history and practices of the speakers to determine the status of the variables and to provide an explanation for
the patterns we find.
What have these challenges taught us about the ways we should view social patterns of variation and change? Can
we still make generalisations? Stratification, accommodation, and differentiation are three basic patterns, and in
some senses, they form general explanations on the dimensions of power and solidarity, or status and friendship.
Some variables display power or prestige. Prestige and power are imprecise words; however, prestige or power
represents some level of worth in a social marketplace, and that such a marketplace is not always accessible to
everyone in the same way. This perspective may carry over too many gender differences, as men and women have
different orientations not just to the economic marketplace, but also to what Eckert calls the heterosexual
marketplace. In other words, a valuable straight man performs social practices, and symbolises them linguistically
differently from a valuable straight woman. Other variants are related to accommodation and solidarity. But this
accommodation is not necessarily automatic. We tend to talk the wat people we like do, and hope that by doing so
we will make other like us. The key to understanding this pattern is to determine which variables and variants are
more or less imbued with market value.
As Dubois and Horvath argue, describing patterns in terms of these categories is not an explanation. Explanations
for social patterns come from understanding what variants symbolise for different speakers in a speech community,
how history, ideologies, and practices imbue the variants with meaning, and why a variant is a desirable way of
speaking for those speakers who use it. So these patterns trace people’s and communities’ lives and the explanations
will need to relate to these lives, and their everyday use of variation. What we need, then, is to look more into the
process of how speakers use variables to create social meanings in their daily interactions.
As we search for patterns of variation and change in the speech community, one of the challenges is how to
represent the fact that individual speakers use different ways of speaking in different situations. We have considered
speakers in speech communities to be single ‘data points’, often with one value for the rate of use for a particular
variant. However, there are no speakers who speak exactly the same way all the time.
Labov was concerned with intraspeaker variation. Labov shows that we can correlate measures of attention paid to
speech with rates of use. This within-speaker variation is an important fact that variation studies must address, for
a number of reasons. First, we do not want to compare ‘a casual salesman to a careful pipefitter’ and conclude that
the patterning of their overall variation is the same. We need to know the effect of both identity and speech event,
and we need to be able to separate the two. The second reason is that the overall patterning can indicate the
distribution of the values attached to the variants in a speech community. Labov argues that it is the prestige variant,
on the assumption that when people pay attention to their speech, that they are likely to shift to. Finally,
intraspeaker variation can provide a window on the linguistic organisation of the speech community: it forms
another dimension of norms about which kinds of speech should be used in what kinds of situations. This
understanding of intraspeaker variation takes us beyond the casual-careful continuum and suggests that variation is
embedded and constrained in the speech community by other kinds of speech norms in addition to identity and
prestige.
Labov’s work focus on objectivity, and his approach to individual variation – which he calls style – is no exception.
As we saw in Labov’s study, there was a correlation between the class pattern found in the speech community and
the direction of style shifting: more ‘careful’ styles were correlated with the highest classes. Since Labov’s
development of this method, however, the sub-field of sociolinguistic style has developed considerably.
Before we address the canonical patterns of intraspeaker variation, we should clear up the ways that the term style
has been used in variationist studies in sociolinguistics. The first is simply intraspeaker variation, as discussed by
Labov: what is the type of speech situation, and how do they shift from one type of situation to another? A more
recent view sees style as being like style in fashion: a set of speech practices (such as variables) that pattern together
to provide a certain kind of style. This latter concept of style as discussed by Eckert as ‘a clustering of linguistic
resources’, such that the clustering leads to some social meaning or even a group identity. These two views can be
combined to some extent, so that we might find that variables pattern together in different speech situations.
Reaction to Labov’s ‘attention to speech’ model of Intraspeaker variation (IV) criticised two main things. First, it
modelled only a single dimension of IV; people change the way they speak for other reasons in addition to how much
attention they are paying to speech. Second, paying attention to speech can have different effects depending on
how speakers are paying attention to it. For example, Schilling-Estes shows that speakers sometimes produce speech
that is more vernacular when they are paying attention to it.
Allan Bell was one of the first to argue for a rethinking and theorisation of style based on whom the speaker is talking
to. He articulated an ‘audience design’ (AD) model following the accommodation theories. In essence, style shifting
in AD is accommodation to the addressee, and to a lesser extent to other participants in a speech event, such as
overhearers and possible eavesdroppers. The usual direction of shift is convergence towards the speech of audience
members. However, speakers may initiate shifts in which they diverge from their audience in order to disassociate
from this audience and converge with a non-present reference group. The primary direction of shift, nevertheless,
is responsive or accommodation. Bell argues that speakers also accommodate to a ‘referee’, or an imagined,
idealised addressee. Finally, he suggests that speakers do not always simply follow the speech of another, but may
initiate a change in the chosen referee.
The ‘acts of identity’ model for IV focuses more on this idea of the referee. This approach is articulated by Le Page
and Tabouret-Keller who argue that: ‘the individual creates for himself the patterns of his linguistic behaviour so as
to resemble those of the group or groups with which from time to time he wishes to be identified, or so as to be
unlike those from whom he wishes to be distinguished’.
This view is similar to the referee design of Bell, but it is more focused on the speaker and their alignment with
different groups who might form referees.
• A language-focused axis, with a formal-to-casual continuum, which implicitly posits that speakers are either
‘naturally’ speaking their one vernacular, or ‘unnaturally’ speaking another variety imposed upon them.
• A hearer-focused axis, with speakers changing their speech depending on their interlocutor, real or
imagined.
• Speaker-focused axis, so that a speaker accomplishes an ‘act of identity’ regarding herself or himself.
In addition to the identity-focused and audience-focused motivations for IV, another way to organise IV is to find
differences in the way speakers sue variables depending on how the speaker perceives ‘what is going on’. This has
been variously conceived of as register, speech activity, speech event, genre and frame. All the terms come from the
insight that talk in a speech community is not an endless, undifferentiated flow. Speakers in the community have a
sense of what differentiates situations, and can articulate to some degree what is appropriate and what is not in
each of those speech situations. This organisation of speech events includes terms to describe different kinds of
speech: formal-informal, casual-careful, standard-vernacular.
All of these terms -register, speech activity, speech event, genre, and frame – represent different ways of noticing
structure or norms in a speech community. A register is a set of connected linguistic practices that are associated
with a group of speech situations in a speech community, or with an institution in a speech community. Registers
might be more specialised, such as the one a doctor uses to talk to her colleagues. We thus might think of the
standard-vernacular continuum as a register continuum to the extent that the speech community recognises these
kinds of shifts in regular ways. The attention-paid-to-speech definition of IV assumes that when people pay more
attention to how they are speaking, they move their register ‘up’ to the more standards. This view assumes that
there is an understanding – an ideology – in the speech community that organises registers in terms of standard and
vernacular. While this organisation of speech may seem obvious, this obviousness is precisely what makes it an
ideology. Other registers can be imagined. In fact, in English churches there is often another register that is even
more formal and sounds more like an older form of English than the standard.
A speech activity is a recognised, repeatable activity in a speech community which has speech as an important part
of its definition. Lecturing is a speech activity that students are familiar with, but aspects of lecturing – such as a long
turn at talk – can be used at times when they are not so appropriate.
A speech event is an instance of an event that is speech-focused. It can be thought of as a particular token of a
speech activity: a lecture, a class meeting, a conversation, etc. The speech event is the focus of an approach to
language and culture known as the ethnography of speaking or the ethnography of communication, which has as its
general goal an articulation of the emic (or insider) knowledge that a speaker needs to have to be communicatively
competent in a language, and thus tries to represent how speakers think about the kinds of speech events and the
language used in them in their speech communities. Common components of the speech event are the setting, the
participants, ends, act structure, key, instrumentalities, norms of language, and genre.
Genre is a tricky concept, partly because the term is used in so many ways in different traditions. In Hyme’s use, it is
very similar to speech activity. He gives examples such as ‘poem, myth, tale, proverb, among others. Importantly, he
argues that ‘genres often coincide with speech events, but must be treated as analytically independent of them’.
Genres thus have formal characteristics that can be transported from one speech to another.
Frames are explicitly centred on the ongoing interaction, and how speakers are orienting to ‘what is going on’.
‘Frame’ is related to the notion of genre and speech event. One of the important facets of approaches using framing
is that they posit that the speech we use is not just a reaction to the frame, but in fact the linguistic choices are sued
by speakers to indicate what they think the frame currently is, or to shift the frame.
Coupland points out that voicing can be the motivation for making different linguistic choices in an interaction. In
interaction, we often voice others, either directly or indirectly.
The point of reviewing all of these concepts is that in each one, various social practices go together: how we sit or
stand, where do we do what, when we talk, and of course how we talk. They thus represent a way that variation is
embedded in the speech community. Note that for all of these concepts, there is a tension between iterability and
creativity. Iterability is the repetition of a form in the same context: that is, in the activity of lecturing, for example,
certain kinds of speech are repeated over and over again. We may even say that there is a community ‘rule’ or ‘norm’
that requires the use of this kind of speech. However, this iterability can also be sued creatively, as when a fragment
of the lecturing activity is inserted into a different speech event, such as an argument.
The concepts used above can be used to code sociolinguistic data, and in the end can point to important patterns of
variation in the speech community. There is little research that explores the variation patters of speech activities,
frames, genres, voices, etc., but there is every indication that these are valid constructs that speakers rely on to
make linguistic choices. The problem is making reliable coding mechanisms and could prove to be a fruitful area of
research.
We have held a number of possible motivations for speakers to make linguistic choices. But what are speakers
actually doing when they make choices (whether conscious or unconscious) about what variant of a variable they
use? Conversations are social practices, because they relate one person to another in some way. This relationship-
indicating function of language is one suggested by many of the patterns we have seen: power (stratification) and
solidarity (accommodation, audience design, and acts of identity) are the most general motivations for speakers to
choose one way of speaking over another. But we have mostly been considering these dimensions of power and
solidarity from a wide perspective. It is more likely that speakers tend to be more concerned with what is happening
now in this particular conversation.
One possibility that has been held by sociolinguistics is whether the topic of an interview or conversation has an
effect. The results here are inconsistent. One problem is that it is difficult to provide a reliable definition of topic,
and that the coding of topic may be such that topics not relevant to the speakers or the variables being measured
are coded for. A study by Schilling-Estes provides an example of the effect of topic. She analysed the use of six
phonological and morphological variables as used by two men from different ethnicities. The variables were all
characteristic of African American English or Southern American English. Schilling-Estes finds that for some topics,
and for some variables, the topic makes a difference, especially in whether the two men converge or diverge in the
use of a variant. She shows that this result is not simply due to whether the men are aligning more to ethnic harmony
or to difference, but that they are orienting to each other and the ethnic categories in complex ways. Thus, while we
may find that changing topics is a way to find quantitative patterns in interactions, it is still a measurement rather
than an explanation. It is how a speaker relates to a topic, and how that topic mediates the relationship between
interactants, that are the engine driving the variation patterns. Schilling-Estes thus also suggests that what speakers
are doing in conversation is creating relationships in their talk simultaneously with the topic of the talk and the
people in the interaction.
A term increasingly used to discuss the creation of relationships is stance, which can be defined in a number of ways.
For example, are we taking an aggressive stance or an uncertain one? We will define stance in two ways: as a person’s
representation of their relationship to their talk (their epistemic stance – e.g., how certain they are about their
assertions), and as a person’s representation of their relationship to their interlocutors (their interpersonal stance –
e.g., friendly or dominating. Epistemic and interpersonal stance are often related: someone who is being patronising
(interpersonal stance) is usually expressing the fact that they are very certain (epistemic stance) about what they
are saying.
We can use variationist methods to show what kinds of general stances a person takes when using a linguistic variant,
and we can use qualitative analyses to understand and interpret them.
Stance can be used to explain a number of the IV patterns we have already seen. For example, Schilling-Estes shows
that it is not the topic of conversation that drives the divergence or convergence of her two speakers, but the stance
they take to that topic, and how it relates to or mediates their interpersonal relationship. As their stances to the
topic of race relations change, then, so do their stances to each other, and this change in stance is partially
constructed through the rates of variation.
One of the most important studies in the audience-design and accommodation literature is Coupland’s study of one
travel agency assistant called ‘Sue’ in Cardiff. First, he coded Sue’s talk into four ‘contexts’, which were very much
like speech activities: casual, informal and work-related, client, and telephone contexts. He measured Sue’s variation
in five variables that had local and more ‘standard’ variants, and found a clear pattern of using the local variants the
most in casual contexts, and decreasing through informal, client, and telephone contexts. Here we could argue for
stance as a factor, as one often takes a different stance to people depending on speech activity and who is in it.
Sue accommodates to the client’s wishes by definition as an ‘assistant’. Other instances of interaction, in which
different stances are taken, may not be so ‘accommodating’. So it is partially the ‘audience’ that is important in IV,
but it is the stance that the speaker takes with respect to that audience that is the real explanation for the variation.
Another important study in IV is Rickford and McNair-Knox’s study of Foxy, an African American teenaged woman.
They analyse two interviews with Foxy, one with her daughter Roberta, both African American, and one by a 25-
year-old White postgraduate student. The variables they measure are all associated with AAVE. For all of them, Foxy
uses a higher rate of the AAVE variants with the African American interviewers than with the White interviewer.
However, different topics do not always have the same effect. For example, school topics have some of the highest
rates of AAVE variants with the African American interviewers, but some of the lowest rates with the White
interviewers. But the researchers also note stance differences in the two interviews, characterising the interview as
being ‘livelier and more informal’ with ‘more give and take’. So the interview has an effect on the stance that Foxy
takes.
A number of studies show that linguistic variants are used to create stances without explicitly appealing to the notion
of stance. As an example, let’s take Miriam Meyerhoff’s study of the discourse particle eh in New Zealand English.
She found that Maori men used eh the most, but that young Pakeha women seem to be adopting it. She shows that
eh functions as a ‘positive politeness’ marker. Positive politeness is a set of strategies that show appreciation for an
interlocutor’s self-worth, or ‘positive face’. They are strategies that suggest a solidary stance. She also argues that it
is the Maori men and some Pakeha women who would be likely to value such stances, thus tracing the quantitative
pattern back to the stance functions of eh.
How do variables mean? In other words, when a speaker makes a choice between two variants, would that utterance
have been understood differently if the speaker had made another choice? Of course, according to the original
definition of the sociolinguistic variable – two or more ways of saying the same thing – there is no meaning
difference. But that definition refers to a particular kind of meaning, which is called a symbol and is characterised
by arbitrary, referential, and denotational meaning. In other words, two words can refer to the same thing (like pop
and soda).
The word (signifier or sign) has an arbitrary or conventional connection with the referent or object. This is the kind
of meaning we are used to thinking about. Another type of sign is the icon, which is a sign that resembles its object.
A common example of this is a traffic sign. In spoken language some words are iconic, in the sense that they sound
like what they refer to, such as buzz referring to the sound made by a bee.
Finally, there is the category of sing called an index. This type of sign is one that is inherently or directly connected
to its referent. A non-linguistic example is a human footprint. In language, deictic words such as this/that, here/there,
and now/later are indexical because their meaning is directly connected to where the referent is with respect to the
speaker (this/that), where the speaker is (here/there), or when they are speaking (now/later).
Indexical meaning is the meaning we are after when we are talking about sociolinguistic variation, although there is
an element of symbolic meaning as well. Let’s take the pronunciation of /r/ in English as an example. If we hear an
/r/ articulated in a constricted, alveolar way, there are a number of ways we might interpret it. If we live in New York
City, we have probably heard lots of people pronounce /r/ this way, and we may remember something about those
people. The next time we hear /r/ pronounced this way, however, this memory surfaces as a kind of indexical
meaning for /r/, because we make an inference that there is a connection between how a speaker pronounces /r/
and other facts about them. We have direct indexical experience.
We can also have a mediated indexical experience, in which case the meaning of the variant takes on aspects of a
symbol. By mediated I mean that someone points out to us the relationship between the variant and its indexical
meaning. Ways of speaking and who uses those ways of speaking may be noticed by speakers in a speech community,
and they become more and more conventionalised. Even if I have never been to New York, I may have heard
someone discussing /r/-pronunciation there, and if I later visit I may already have formed an opinion that a vocalised
/r/ is more working-class.
Labov used the terms indicator, marker, and stereotype to discuss this distinction:
• An indicator is a variable that has a social pattern, but one that the speakers are not yet aware of. That is,
some people speak one way, others speak another way, but no one in the speech community has yet
noticed the difference. Moreover, Labov defines an indicator as showing no regular IV pattern.
• A marker shows IV shifts, and there is a subconscious awareness of the social group patterns of language.
• A stereotype is a variable that everyone in the speech community is aware of and can talk about.
Michael Silverstein brings these ideas and insights into a more general theoretical construct, which he calls the
indexical order. The indexical order categorises indexes as to their relationship with metapragmatic function.
Metapragmatics is the use of a language to represent pragmatics, or how speakers use language. Silverstein is
interested not in what is purported to be ‘in the heads’ of speakers, but in how metapragmatic discourse is circulated
throughout the community.
So for a first-order indexical the connection between, for example, a category of speaker and a variant might be
present as a demonstrable correlation, but there is no metapragmatic function in the community that imbues it with
the indexical meaning connecting the way of speaking to the type of speaker. Indexes become second-order as the
metapragmatic function makes the correlation between the type of person and the way of speaking more
meaningful. As the index becomes more fixed into the metapragmatic function, we say that it is becoming
enregistered – becoming a recognised index in the speech community. Finally, there can then develop a third-order
indexical in which the meaningful correlation of the second-order indexical becomes the subject of yet another
metapragmatic function.
There is one more set of ideas to discuss: the distinction between presupposing and creative indexes. An index is
used in a presupposing sense it if assumes that a linguistic form is used in a particular situation; for instance, in the
classroom in front of students I call a fellow professor ‘Dr Paulston’. If I use this form of address in another classroom,
I am following expectations that in a classroom we use ‘title + surname’ with instructors. It is thus a presupposing
use of the address term ‘title + surname’. I can also use this address term creatively, however. For example, suppose
that I am speaking casually with this same colleague at an informal lunch. The corresponding presupposing address
term would be ‘Christina’ in this case. However, if at some point this colleague makes a point that reveals my
younger, more inexperienced status, I may respond by saying something like ‘I understand now, Dr Paulston’, in
which case I am using the ‘title + surname’ address form creatively to index her status as instructor. This distinction
is important because the move from one level of indexicality to another requires repeated and noticed creative
indexing of the linguistic form.
What evidence is there that people attach such indexical meanings to these sometimes subtle differences in
language? Linguists and social psychologists have used experimental evidence to show that meanings are indeed
attached to different ways of speaking. The most common technique in this field is known as the ‘matched guise’
technique, pioneered by Wallace Lambert. In studies focusing on the perception of speakers of French and English
in Quebec, Lambert played Canadian English, Canadian French, and Continental French voices to Quebecers, and
asked them to rate the speaker on a number of personality traits using a semantic differential, a scale representing
antonymic qualities. For example, does this person sound more or less intelligent, ambitious or friendly? They found
that there was a somewhat complex interaction between the language background and gender of the respondent
and the language and gender of the stimulus voice, or guise.
Matched-guise-type studies have been performed with different kinds of sociolinguistic variables or sets of variables.
Labov performed a similar study in Philadelphia based on four vowel variables and one consonantal variable, but
had responses based on two traits only: job suitability and friendliness. The job suitability scale asked speakers what
job the person was most suitable for, with television personality as the highest and no job at all as the lowest. Labov
found a strong reaction when listening to guises with the most advanced tokens (in terms of the changes ongoing in
Philadelphia), speakers downgraded the speaker’s job suitability. Labov argued that these results showed that all
speakers evaluate variation in a uniform manner.
This study points to a danger of the matched-guised method: the experimenter chooses ahead of time the
possibilities for how a speaker may respond. The matched-guise technique is important in that it shows without a
doubt that linguistic variants do index speaker categories, but we must remember that these results derive from a
decontextualized experimental task and realise that the meanings detected are shaped by the very task.
In addition, note that all of the stimulus voices in Labov’s experiment are women’s. Given the different reactions for
men and women found by Lambert, we might wonder what effect gender or other social characteristics of the
stimulus might have on a respondent. A number of studies show that what a listener knows about a speaker before
they hear anything leads to different indexical interpretations of that speaker. Two studies show this particularly
clearly, with each priming participants in different ways. Priming refers to giving the experiment participant
information that may bias their response in the experiment.
Campbell-Kibler’s matched-guise-type studies shed light on how people use variation to create opinions about the
people they are listening to – how variants index social traits. Campbell-Kibler’s work in general shows that other
variables and other ways that speakers think about the speaker affect the way they interpret the indexicality of a
particular linguistic variable.
Campbell-Kibler focused on the (ing) variable in English, and used extracts from sociolinguistic interviews as stimuli.
The general results show that the meaning of a variant is highly sensitive to other factors. One effect was perceived
region, or where respondents thought the speaker was from. Campbell-Kibler found that the effect of (ing) changed
depending on how the respondents categorised the speaker in terms of class and perceived region. If they were
judged to be from the south, they were heard as less intelligent overall. However, if they were not heard as being
southern, and they were judged to be working-class, the (ing) had a large effect. Of course, (ing) plays a part in the
judgment of class and region, but the matched-guise test shows that the (ing) variant switch makes a significant
difference only in this case.
The complexity of Campbell-Kibler’s work is that it shows that while there may be some vague central meanings to
different variants, they are realised only in particularly contexts and that these contexts interact to create a unique
meaning in each interaction. Of course, it also means that previous interactions with a speaker or in a particular
situation will colour our interpretation of social meanings of the same particular variant. Most generally, however,
the perception taken together show that we can detect indexical meanings beyond the correlations between variant
and identity category or other social variables.
Johnstone et al. show that (aʊ) becomes a third-order indexical when the speech community comes into contact
with exonorms. Recall that Dubois and Horvath showed how some Cajun variants lost favour in Louisiana, and then
reversed as the culture began to value Cajun social practices, including some Cajun ways of speaking. The non-Cajun
variants were in this case exonorms, and their appearance moved some of the Cajun variables from first- to second-
order indexicality. The polarity of this indexicality was for the most recent (male) generation reversed such that
Cajun-ness became valued, and these variants then moved towards being third-order indexicals. In other words, the
Cajun way of speaking was in a sense ‘re-enregistered’.
This indexical cycling of enregistrations and tensions among norms are central processes in sociolinguistic variation
and change.
• For accommodation, the first possibility is that there is no indexical value at all, and people follow a
predisposition to mimic the talk they encounter. However, accommodation may also come into play when
there is an indexical value, for example one in which there is a standard-vernacular metapragmatic
relationship.
• Differentiation is something that will generally occur for variants with indexicality at the second-order level
at least. The avoidance of /au/-monophthongisation by African American Pittsburghers shows that they
associate monophthongisation with White speakers, and it is clear that this variable has reached third-order
status.
• Stratification is dependent on not only the metapragmatics of the speech community, but also they way
groups are valued and organised in that speech community. For working-class speakers in Pittsburgh in the
early 20th century, monophthongisation was a first-order index. However, as the variable moved up the
indexical order, its indexicality also acquired the values attached to the identities with which it was indexed,
and fitted into a more basic metapragmatics in which there is ‘good speech’ and ‘bad speech’. We find in
stratification, then, the conflicting orientations to the metapragmatics of ‘standard language’ with the
upper class oriented more positively and the middle and working class oriented (mostly) negatively.
‘Values attached to language’ have also been referred to as language ideologies, which are beliefs about language
and its use that are taken as natural and not open to question. That there is a ‘correct’ way of speaking is itself an
ideology, largely tied up with ideologies about class. Lippi-Green identifies this as ‘standard language ideology’.
Although standard language ideologies are the most obvious, we can find alternative ideologies and enregistrations,
such that, for example, using RP is to be avoided so that one is not perceived as being too posh.
The meanings of a variant are thus intimately tied to the social structure and cultural values of the speech
community. We find that there is a repeated change of meaning of variants, sometimes in concert with social change,
but sometimes being recruited to support the status quo in a new way. Moreover, work in perception and in studies
on stance shows that the indexicalities of variants are not only not fixed, but vague and contestable. How can we
present, then, the way that choices of variant make social meaning?
As Silverstein points out, these indexicalities are always both creative and presupposing: they rely on previous use
for meaning, but are being used in a new situation and thus creating new indexicalities. Even when used in a way
that exactly follows the ‘rules of use’, indexes reinforce that meaning and are thus re-creating. Eckert argues that
we should think of the indexicalities not as a single meaning, but as an indexing range of potential meanings which
calls an indexical field. Importantly, however, ‘this field is not a static structure, but at every moment a
representation of a continuous process of reinterpretation’.
Can we represent the connections between forms and indexicalities? Can we represent connections among
indexicalities? In creating meaning, speakers are likely to use associations among the indexicalities to create new
meanings: for example, a person who sounds educated could also sound more articulate since there is an ideological
connection between these two.
Ochs proposes a similar process in which indexes are indirect. She argues that gendered particles in Japanese
originally had stance indexicalities such as rough and refined, which were then indirectly indexed to the genders that
were rough and refined. The connections among non-linguistic indexes we can more generally call secondary
indexes, with primary indexes being the connection between the linguistic item and the concept. Of course, we may
not always be able to tell which the primary index is, but it is important to realise that by indexing one quality or
stance of a person, we are inserted into a complex web of indexical meanings that can be foregrounded or
backgrounded in actual use. there is thus an interaction between the indexicalities in the relatively more durable,
but not unchangeable, indexical field and the meanings created in interaction.
Let’s take a brief look at some of the actual indexicalities that variants take on, and the semantic relations among
them. Identity has been seen as a motivation for variable choices. Labov showed that speakers with a more English
islander identity were more likely to use centralised diphthongs than those who were more oriented to the mainland
or less oriented to the English identity. We have also seen the ‘acts of identity’ model for style, in which speakers
identify with certain groups by using language associated with that group.
Eckert analysis of the Detroit high school has set the template for a number of other studies that have similar general
findings. Eckert’s work revealed that the social world was ideologically organised around the opposition between
jocks and burnouts and by gender. The jock and burnout categories were defined by the social practices engaged in
by the students and secondarily by their social networks. The important point is that the jock and burnout labels
were ideologically recognised and practice-based categories that were indexed partially by the different vowel
variables.
Eckert’s work has inspired a number of studies that elucidate further the details of how locally relevant, practice-
based, and ideologically recognised personae are related to larger patterns of variation. All of these studies show
how variation is used by speakers to align themselves with personae in communities of practice which can be
understood as local ways of experiencing larger social categories such as class and gender. These personae are in a
sense made up of stances, such that there is a secondary indexicality between the personae and the stances that
the speakers take up. Bucholtz and Hall refer to this process as stance accretion.
We this have three abstract levels of social organisation that are related dialectically and interact to produce
indexical meaning and change in such meaning. The largest is that of the groups traditionally coded for by
variationists (age, gender, class); the middle level is that of communities of practice and the ideologically relevant
personae that they are inhabited by; and finally there is the level of stance. These levels interact in both the
interactions in which the variables are used and over longer stretches of time as a change expands through a speech
community.
This three-level view gives us a fairly good understanding of how variation and change are embedded in the speech
community and spread throughout it. The actual path of a change is still very much a mystery, but it looks as though,
as a change becomes older, its meanings become more diffuse and vague. New generations will also reinterpret the
meanings, and they will expand as they are used in wider contexts. As well, in order for speakers to create a
distinctive persona, they will need to advance a change further.
ACQUISITION OF VARIATION
HOW IS VARIATION LEARNED?
Little work has been done on exactly how children acquire variation, and what patterns they show as they develop
their linguistic systems. Labov proposes a model of transmission that has language learning in the early years – from
the caregivers, who he notes is usually female – as central, so it is important to uncover the processes and constraints
in this transmission process.
EARLY CHILDHOOD
Children’s production from early on is necessarily variable, in fact more variable than adult speech. Much of this
variability, however, is probably part of the learning process as children refine their linguistic systems and in fact
narrow the variability.
One of the challenges of studying childhood is determining what part of the variation is developmental – related to
the process of acquisition – and what part is part of the system they are acquiring. However, little work of this sort
has been done.
It is clear from the research so far that children learn variation as an integral part of the language acquisition process.
The earliest age that has been reliably reported is about three, and the studies that have been performed show that
some fairly complex patterns of variation already appear. A classic early work in in this are is Payne’s. She found that
early exposure to phonemic categories is crucial in acquiring this system.
Roberts analyses the variation patterns for coronal stop deletion (CSD) among 3- and 4-year-old children in
Philadelphia, and finds that they have mostly acquired the following environment and morphological constraints.
Labov shows that the structural constraints on CSD and (ing) are already present in the speech of a 7-year-old boy
in Philadelphia, and also shows that the relevant constraints are present.
Kerswill, however, working in a dialect contact situation, shows that fairly young children are active in creating the
new dialect. In other words, they are already deviating from the language of their caregivers and towards a new
community norm. This result suggests that even at a young age, if children have contact with other children, these
peers will have some influence on the development of variable patterns. However, a later analysis shows that the
youngest (4-year-old) children have the closest correlation with their caregivers, while older (8- and 12-year-old)
children correlate less with their caregivers, showing that there is an influence of caregivers that is greater at the
younger stage than at the older.
One of the most important recent studies in this area is Smith et al.’s study of 2-to-4-year olds. They discover how
two different variables are acquired in different ways. They show that the structural constraints on both variables
are present in children from a young age. We might predict that in general, the more regular constraints and patterns
seem to be adopted at an earlier age, while more complex, specific, and abstract patterns show up later; but this is
not always the result.
Smith et al. also show that the caregiver has a different effect on the children’s use of each variable. They found a
very significant effect for both structural and social patterns in one of the variables, but not in the other. So for some
variables caregivers adjust their child-directed speech, while for others they don’t. It is a reasonable hypothesis that
it is related to indexical status (first- or second-order) of the variable.
In addition to the structural patterning at a young age, the social use of variation is also present at this stage. Labov
shows regular intraspeaker patterning for CSD and (ing) for a 7-year-old in Philadelphia. Such patterning is difficult
to detect, since most young children’s speech is not collected in extended interviews. Smith et al.’s study in this
regard is again important. They found that caregivers have a regular intraspeaker shift between roles in which they
are creating a clear authoritative persona (teaching and discipline) and ones that create less difference between
them and the child. Most importantly, the children match this difference exactly. This result hints that the early
acquisition of social variation is related to the different prototypical roles played by parents and children. This
pattern is a view into the dialectic between the pattern of the wider community and its sue as a stance index, and
shows how the indexical relationship in the community are recreated in new generations from early on.
So in terms of both structural and social constraints, and age of acquisition, variables differ considerably. It is not
clear at this point what is important in determining these differences. Some possibilities are the level of abstractness
or complexity involved in the variable, the frequency with which tokens of the variable are present, the indexical
order of the variable, the difference in rates of caregiver use, and how that is related to the indexical order.
Older children have been much less studied than both young children and adolescents. However, it is clear that his
stage of life is important. This tends to be the age range (about 6 to 12 years old) when children spend less time with
caregivers and more with peer groups.
Kerswill and Williams do include 8-years-old (along with 4-year and 12-year-ikds) in their sample in Milton Keynes.
They find that for their (ou) variable, the youngest group is correlated with caregivers, while this correlation
disappears for the 8- and 12-year-olds. This result provides clear evidence that it is at this young stage that children
begin to move to the variation norms of their peer group, and that this is probably where the incrementation of
change begins to take place.
Eckert shows that it is at this stage that children being to enter what she calls the heterosexual marketplace as they
build a ‘peer-based social order’. This social order “emerges in the course of the fifth and sixth grades in the form of
a heterosocial crowd, which is an alliance of smaller friendship groups of kids who have emerged as popular through
early elementary school”.
Eckert shows how this heterosexual market becomes a linguistic one, and how linguistic variants become involved
in the kinds of stances that girls and boys take in order to be successful on this market. It is clear from this work that
at this point, the ‘kids’ are not so much reproducing the variation that they learned from their parents as using that
variation as a resource for social work in this crowd. Her work shows that this stage sees the beginning of the
emergence of each generation’s adolescent contribution to language change. Eckert’s groundbreaking study of peer
groups and variation showed how these changes are embedded into the social structure and practice of adolescents
in high school. Central to that practice is the creation of what Eckert calls a persona. As one group moves to create
the most distinctive stylistic practice, they increment changes further away from more ‘conservative’ speakers.
These speakers don’t experience themselves as incrementing a change, but as aligning themselves with certain
stances and personae that are indexed by the most extreme variants, and in the process increment the change.
ADULTHOOD
Adulthood is often not thought of as a developmental stage, but it is, and in fact one can identify different adulthood
stages as well. However, linguists have often viewed adulthood as the end-point of acquisition. While this may be
the case for some variables, it is clear that changes do take place over the course of speaker’s adult lives.
There are also likely to be changes in adulthood as adults move into new speech situations and roles. For example,
moving from being an unmarried 25-year-old to a 35-year-old parent changes the variable patterns a person uses,
and what variables are sensitive to such a change.
This hypothesis seems sound at a very general level: some variables are differentiated by style early on, and children
match their caregiver’s pattern. However, we don’t see all variables being learned this way; some are more complex,
or they are not used by the caregivers to differentiate stance or speech activity. But it is clear that some variability
is learned by children from caregivers.
It is what they do with that variability as they grow up and then grow old that is of central interest to a theory or
model of change. For we have seen that children actively learn and create new ways of speaking in their peer groups.
In most cases, this appears to be a mechanical process in that many people move in the same direction. However,
this incrementation seems to happen because the variability is a resource for social meaning locally and individually.
The term ‘acquisition of variation’ is thus problematic, because at first children acquire only a general variation
pattern; later on, they actively create their own patterns of variation, thus moving a change forward, or reversing or
recycling it.
UNIT 4 GLOSSARY
Accommodation > It is the process by which speakers modify their own speech to the one used by the person who
is interacting with them, whereby one group of speakers show solidarity with other speakers to facilitate
communication. This way of adjusting one’s own speech can give way to speech convergence or speech divergence
depending on the intentions of the speakers and the results of the communication encounter.
AAVE (African American Vernacular English) > It refers to the languages spoken in black communities in the United
States. Some linguists consider it a significantly different linguistic system from the standard dialect since it does not
conform to its pronunciation, grammatical structure, etc.
Change by diffusion > It is a change from above in which speakers are conscious of and which is characterised by a
spread of one variant from one geographic are to another, mostly by adults.
Change by transmission > It is a change from below in which a change is under way and is moved forwards by
children as they acquire a variety.
Differentiation > It is a situation in which a group of speakers change their variable usage so that it is different from
another group. Speakers want to contrast their identity with a certain group.
Ethnography of communication > It studies the norms and rules for using language in social situations in different
cultures. It deals with aspects such as the different types of language to be used under different circumstances. It
includes non-verbal aspects of communication, for instance, distance between speaker and hearer, eye contact, etc.
Frame > Frames are explicitly centred on the ongoing interaction, and how speakers are orienting to ‘what is going
on’. ‘Frame’ is related to the notion of genre and speech event. One of the important facets of approaches using
framing is that they posit that the speech we use is not just a reaction to the frame, but in fact the linguistic choices
are sued by speakers to indicate what they think the frame currently is, or to shift the frame.
The basic idea is that one cannot understand the meaning of a single word without access to all the essential
knowledge that relates that word. One would not be able to understand the word “sell” without knowing anything
about the situation of commercial transfer, which also involves a seller, a buyer, goods, money, the relation between
the money and the goods, etc. Therefore, a frame would be a collection of facts that specify “characteristic features,
attributes and functions of a denotatum, and its characteristic interactions with things.
Genre > Genres are types of spoken and written discourse recognized by a discourse community, for instance,
lectures, conversations, speeches, etc. Each genre has typical features. Also as part of the ethnography of speaking
framework, it is a tricky concept, partly because the term is used in many traditions: poems, proverb, riddle, etc.
Genres often coincide with speech events but must be treated as independent from them.
Gravity model > It is a way of geographical stratification. In this case, the diffusion takes place from the most
populous cities and then to the second-most-populous cities, skipping over very small towns.
It predicts that the larger the city/town, the sooner an innovation is likely to show up there. It is a way of
geographical stratification. It indicates a pattern in which the urban centre, which often has the most prestige and
cultural capital, is in a way equivalent to the upper-class speakers.
Hypercorrection > It is a manifestation of linguistic insecurity, for instance, in a social group. It can manifest itself by
the overuse of socially desired forms in careful speech or reading, especially in an attempt to speak or write in an
educated manner. For instance, a speaker of a non-standard variety of English may practice more self-correction
when speaking formally and makes use of more sophisticated vocabulary or a clearer pronunciation.
Incrementation (of a change) > It is the advancement of a change in the same direction by the next generation of
speakers.
Interspeaker variation > It is the differences and variation that is measured between different speakers (individuals
or social groups.
Intraspeaker variation > It is the differences in the way a single person speaks at different times, or within different
interlocutors, or even within a sentence.
Koines > They are generally discussed as new dialects that are formed from mixture of previous varieties. When
large groups of people come into contact, the members of the groups end up speaking in ways that are similar to
each other and eventually form new varieties known as koines.
Patterns of interaction > These are the different ways speakers usually interact. Using the right interaction pattern
is a fundamental factor in the success of any interchange between speakers.
Register > It is a set of language practices that are associated with a group of speech situations in a speech
community, or within an institution in a speech community.
Speech activity > It is a recognised, repeatable activity in a speech community which has speech as an important
part of its definition. Lecturing is a speech activity that students are familiar with, and one that often happens in
classrooms.
Speech community > It is a group of people that share the same language or dialect in a specific setting which can
be close, such as a city, a neighbourhood or a whole country. Speech communities not necessarily correspond with
political boundaries, religions or cultures.
Speech event > It is an instance of an event that is speech-focused. It can be thought of as a particular token of a
speech activity: a lecture, a class meeting, a conversation, etc. The speech event is the focus of an approach to
language and culture known as ethnography of speaking or ethnography of communication.
Stratification > It is the process in which language is used differently by groups who are placed on an ordered
prestige scale. It represents differences in the prestige attached to variants. Socioeconomic class is the prototype of
stratification. This variant reflects hierarchical or power relationships in a given speech community. Societies decide
which language varieties are prestigious and which ones are not.
Structuralism (in sociology) > It refers to a view of society as having a structure of distinctions in a system. The terms
are only meaningful in relation to their structural relationship to the other terms of positions in society. Being a man
means not being a woman, for instance.
Style > This implies a specific choice on the part of the speaker: formal language for solemn events, less formal
language for everyday situations, and informal and casual language for trivial conversation or relaxed matters.
Transmission (of a change)> It refers to the passing of change from one generation to the next.
Wave model > It is a way of geographical stratification. It diffusion takes place gradually over geographic space, that
is, it emanates from a single starting point and is gradually incorporated into the speech of the nearest neighbours.
As the gravity mode, it indicates a pattern in which the urban centre, which often has the most prestige and cultural
capital, is in a way equivalent to the upper-class speakers.
1. LANGUAGE CONTACT
Language contact studies variation between different language varieties (both languages and dialects) and the
influence they receive from each other. The main objective is to describe language as both a stable and a changing
reality. Some of the disciplines which study language contact are Sociolinguistics, Social Psychology of Language, and
Applied Linguistics.
One of the main reasons why language changes is that they come into contact with other languages. This contact
typically involves bilingual speakers, i.e., people who speak the two (or more) languages involved. The languages of
such individuals may act upon and influence each other in a wide range of ways (e.g. in the adoption of features of
pronunciation, the borrowing of words, or the modification of grammar).
It can be said that virtually all languages have been influenced at one time or another by contact with other
languages. For thousands of years, human beings felt the need to communicate with other human beings for the
purpose of the exchange of goods and trade. It was very common for a few members of the tribes/communities in
contact to acquire some knowledge of each others’ languages.
Whenever people speaking different languages come into contact, there is a natural tendency to find ways to
communicate with very different results: from vocabulary loan to the creation of new languages. These possible
outcomes depend not only on linguistic but also on social and psychological factors.
There are several large contact languages (lingua franca) in the world. English is at present the most widespread
contact language. Formerly French was also a widespread contact language, but it has now lost its role to English.
The consequences of language contact have been studied since the 19th century. For over a century the mainstream
of Historical Linguistics focused on the motivations and internal mechanisms that produce language change, as the
Comparative Method, which assumed that virtually all linguistic changes arise due to intrasystemic reasons. No one
took into consideration mixed languages, since it presupposed influence of other languages. However, the study of
Pidgins and Creoles led to other studies on language changes that result from language contact.
In the 20th century, Weinreich and Haugen emphasized the importance of language contact from both a linguistic
and a social perspective. Contact linguistics has become a subdiscipline of Linguistics that in recent years has been
reinforced by the work of Thomason and Kaufman in which they study various phenomena of language change and
try to lay the groundwork for the typology of the outcomes of language change.
Thomason and Kaufmann object to the structuralist view since they believe that the history of a language is the
history of its speakers and not an independent phenomenon that can be studied without reference to the social
context in which they are immersed. Likewise, they oppose the linguistic constraints postulated by the linguists of
the Prague school or Weinreich himself who maintained that language structure defines or decides what can happen
due to external influence. According to Thomason and Kaufmann, these assumptions are invalid, and argue that any
linguistic feature can be transferred from one language to another. The most extensive case study carried out in
their book is the one related to English in contact with other Germanic languages, in an attempt to establish the
types of linguistic changes caused by language contact in the Old, Middle and Early Modern English periods.
Historical linguists study language caused by contact between languages or dialects. In dialect contact primacy is
given to the social process, while in language contact primacy is given to linguistic factors. However, in both dialect
and language contact, linguistic and social factors must be taken into account in any explanation of linguistic change
as dialects of the same language may have more different structural points than analogous structures in languages
genetically related. Moreover, in many language communities contact with other languages is more frequent than
contact with other dialects geographically distant.
Most linguistic processes that take place through language contact are the result of bilingualism situations in which
there is a need to communicate with speakers of languages that are not mutually intelligible.
• Those involving language maintenance > It refers to the preservation by a speech community of its native
language from generation to generation.
The language can change but only by small degrees and owing to internal developments and/or (limited)
contact with other languages.
• Those involving language shift > It takes place when contact with other languages can lead to the partial or
total abandonment of a group’s native language in favour of another. In some cases, the minority language
may preserve its L1 for certain functions, while acquiring the dominant language for other uses.
• Those that lead to the creation of new languages > Under particular circumstances, intensive language
contact may result in the birth of new types of contact languages, new languages which include bilingual
mixed languages, pidgins and creoles.
• Borrowing > the incorporation of foreign features into a group’s native language by speakers of that
language.
• Convergence > the process whereby languages become structurally more similar to each other.
The study of language contact has given way to other linguistic subfields, such as Pidgin and Creole Linguistics or
Creolistics. The study of pidgin and creole languages has accelerated since the 60s and 70s thanks to studies of
variation and social groups contact, and it is often defined as a subfield of Linguistics within the theory of language
contact. In turn, within this discipline there are others such as, for example, Applied Creolistics.
One of the pioneers in this field of study was Hugo Schuchardt who in the late 19 th century showed several arguments
against the theory that languages do not mix, and against traditional ideas about the genetic relationship of
languages which allowed little change induced by language contact. Schuchardt presented numerous examples of
changes resulting from language contact in bilingualism situations, in the formation of Pidgins and Creoles, etc. This
fact represented a challenge to the Stammbaum theory and the idea of one parent language for all languages. It was
concluded that, although the genetic model is valid for most languages, many others have experienced changes
induced by contact at some point in their history. There is a large number of mixed languages whose subsystems
cannot go back to one single language, as is the case of Pidgins and Creoles.
The advocates of the family-tree model accepted the apparent splitting of daughter languages from their parent
language. However, in actual observation, no speech community is completely uniform. Some scholars did not
accept the Neogrammarian opinion that sound change is regular and exceptional, but rather opposed this and the
family-tree model. The alternative to the family-tree model was the ‘wave theory’ usually attributed to Johannes
Schmidt and Hugo Schuchardt. According to this theory, linguistic changes spread outward concentrically like waves,
which become progressively weaker with the distance from their central point.
Of the approximately 6000 living languages in the world, half have less than 10.000 speakers and a quarter less than
1.000. This amount of small languages has led to widespread bilingualism and multilingualism and many large
languages are used as contact languages.
1.2.1 Pidgins
Pidgin languages developed in situations of minimal and rudimentary communication requirements between two
groups of people, one usually superior to the other in economic and political power.
A pidgin language normally has in its initial stage a limited vocabulary containing words from the language of the
dominant group, usually with a pronunciation reflecting the language(s) of the locals, plus words from the
language(s) of the latter, and a simple grammatical structure, with mostly only short sentences, which reflect
features of the language(s) of the locals. If that pidgin language continues to be used for a long time it becomes more
and more sophisticated.
A pidgin language can become the children’s first language. Such a language is referred to as a creole which has
developed from an original pidgin language through the process of creolization.
By definition, pidgins have no native speakers and it is always a reduced language. The grammar is always simpler
than the grammar of the source language. Most European trading pidgins were based on the languages of the
seafaring nations.
1.2.2 Creoles
European colonial expansion during the 15th to 19th centuries led in many cases to the creation of new communities
of people that were brought from distant regions of the world. For instance, slaves were brought from West Africa
to the plantations of the New World. Contact between them and European settlers led to the emergence of Creole
languages. The term ‘creole’ refers to the locally born descendants of slaves in the colonies.
In the process of creolization, the simple structures of pidgins get elaborated in a variety of ways: morphology and
syntax become more complex, the vocabulary increases, and pronunciation becomes more stable.
Some examples of Creoles include Jamaican and Guyanese Creole, Haitian Creole, and Curacao. Creoles also
emerged in the Indian Ocean where European colonies were stablished.
The term ‘language death’ applies to the extinction of a language, its complete disappearance without trace, or with
traces only in recorded form. Languages can cease to be spoken for two reasons:
• A language dies because its speakers do for various reasons: disease, natural disaster or genocide
• The second kind of language death has to do with intensive language contact. Once group in contact with
a second group disappears because it makes the shift to the other group’s language. The reasons can be:
o Positive: marriage, economic advantage, etc.
o Negative: imbalance of economic and political power
Languages can be endangered due to contact with one or more dominant languages, normally after an extended
period of bilingualism, which threaten the existence of the minority languages, in many cases leading to their
disappearance. People survive but their language does not. Some contact languages have become endangered and
some even extinct. The reasons for this are often oppression by monolingual speakers of dominant languages.
2. DIALECT CONTACT
The difference between languages in contact and dialects in contact is typically defined as involving contact between
non-mutually intelligible as opposed to mutual intelligible varieties.
About 700 million people have English as their first language, and in total as many as 1500 million may be fluent
speakers of English. English is the majority language in Britain and USA, but also in Australia and New Zealand, and
a national language in both Canada and South Africa. In other countries it is a social second language and in others
(Pakistan, Ireland, or the Philippines) it is an official language or the language of business.
All these modern varieties of English derive from the Germanic language-variety spoken by the Anglo-Saxon invaders
who came to England in the 5th century A.D. from the lands across the North Sea. They spoke a dialect of the
Germanic branch of Indo-European which eventually became a distinct language, English, which developed in the
British Islands up to the 16th c. During the centuries of imperial expansion, English was taken to other parts of the
world.
The native inhabitants of the island were Celts, speaking various forms of Celtic. With the departure of the Romans
from Britannia in the 5th century, Germanic tribes settled on the island. Germanic is the term used to describe a
group of languages which share a set of characteristics unique to them some of which are still observable in Present-
Day English (PDE). Old English (OE) was a West Germanic language together with Old High German, Old Frisian and
Old Saxon. The most important present-day languages which are of Germanic origin are English, Frisian, Dutch,
German and Danish, among others. Germanic, in turn, belongs to a larger language family known as Indo-European
which also includes Indic, Greek, Romance, Slavic, Baltic and Celtic, among others.
English has changed over the centuries and it has been influenced by other languages such as the Scandinavian
languages spoken by the Viking raiders who invaded England between the 8 th and 11th centuries, the French spoken
by the Normans in the 11th century, or Latin, which was the language of the church for centuries. Therefore, when
studying the history of the English language, it is important to know the languages English has been in contact with.
Languages are not static, they are constantly changing. Old English is very different to Middle English or to Present-
Day English. Various periods are recognised in the history of English. These periods are characterised by certain
historical, language-external events:
• Prehistoric Old English (approx. 450-650/700) is the period before written records
• Old English (approx. 700-1066). Most Old English which has come down to use is written in the West Saxon
dialect, since the power was centred in the South-Western kingdom of Wessex. However, there is evidence
for at least three other dialect-groupings: Old Mercian, Old Northumbrian and Kentish.
• Middle English (1066-1476). It is marked by the Norman Conquest and the end by the coming of printing in
Britain.
• Early Modern English (from 1476 to the early 18th century)
• Later Modern English (from the early 18th century to the present day)
The Anglo-Saxons had been living in England in relative peace for some three centuries when they began to be
disturbed by their northern relatives. Old Norse belonged together with Anglo-Saxon to the same branch of the
Germanic family. Some major linguistic changes due to contact with Old Norse were:
Some important issues when studying the contact of Old Norse with Old English are: did the Anglo-Saxons learn to
speak Norse or the Scandinavians learn to speak English; or where there something of both? If so, was such
bilingualism widespread?
Some scholars have suggested that there was simplification of the Old English grammar through contact with Old
Norse. They believe that inflexions were non-functional in communication between speakers of the two languages,
and the consequence for English was the loss of inflexions and the development of a fixed word-order, observable
in Middle English. This has led some scholars to maintain that early Middle English can be regarded as a creole arising
from Anglo-Norse contact.
There is a three-century gap between Northern Old and Middle English, since no texts are found until the 13 th
century. However, there are patterns of regional continuity in these varieties, especially at the morphological level.
This, and the fact that the morphological simplification of Northern English started before the Scandinavian influence
could be felt in the North, seems to go against the creolization hypothesis.
UNIT 5 GLOSSARY
Bilingualism > It is the phenomenon of speaking and understanding two or more languages. The term can refer to
individuals as well as to an entire society. It refers to a situation in which two languages coexist in the same speech
community or social group.
Borrowing > It is the process by which words or phrases from one language or dialect is adapted for use in another.
It can be manifested at the level of phonology, morphology and syntax. Borrowing can be originated by close contact
between two or more languages, a sense of need or prestige, or the domination of some languages by others.
Code mixing > It occurs when the interlocutors from one language to the other in course of a speech mix two or
more language or language varieties. Code mixing is relatively common in the speech of bilinguals and immigrants.
Code switching > It is the alternation between varieties, or codes, across sentences or clause boundaries in the
context of a single conversation. It is often used as a cover term including code mixing as well. Tag-switching,
intersentential switch (between sentences) and intrasentential switch (within a sentence) are different types of code
switching.
Contact-induced language change > It is any linguistic change that would have been less likely to occur outside a
particular contact situation”. In these changes, foreign material is transferred from one language to another.
Contact language > It is a language which is usually used by speakers who have different mother tongues and
therefore, need a common language to communicate among them. The most remarkable example nowadays is
English, which is spoken by some people as a mother tongue, many others use it as a second language, and still
others as a foreign language, but, as a rule, it serves as a lingua franca for international and intercultural
communication.
Convergence > It is the process whereby languages become structurally more similar to each other.
Creole > A creole language, or simply creole, is a stable natural language that has developed from a pidgin, becoming
nativized by children as their first language, with the accompanying effect of a fully developed vocabulary and system
of grammar.
Creolistics > It is the scientific study of creole languages and, as such, is a subfield of linguistics.
Creolization > The process by which a pidgin becomes the first language of a group of speakers, when it is acquired
as a mother tongue by children who are exposed to it. The social circumstances in which this language is now used
are more complex as they need to cover all kinds of communicative functions.
Decreolization > It arises when one creole has a prolonged contact with a standard language in a specific society,
and that standard brings a considerable influence to bear on the creole. Speakers start to develop the creole and a
continuum is created with the standard as a model at the top and the creole as a model at the bottom. The different
forms of the creole become socially stratified and the variety or varieties which are closer to the standard often
become the language of the elite and educated society (acrolect), whereas the variety which is closer to the creole
often represents illiterate people and lower social class (basilect)
Dialect contact > It studies variation between mutual intelligible dialects and the influence they received from each
other. Primacy is given to the social process.
Dialect continuum > It is the range of dialects spoken across a large geographical areal, differing only slightly
between areas that are geographically close to each other, and gradually decreasing in mutually intelligibility as the
distances become greater.
Diglossia > A situation where two closely related languages are used in a speech community, one for High functions
(education, church, newspapers) and one for Low functions (home, friends…)
Haitian creole > It is one of Haiti’s two official languages, along with French. It is a creole based largely on 18th
century French.
Language death >It refers to the extinction of a language. Languages can cease to be spoken for two reasons: 1)
because its speakers die due to various reasons (natural disaster or genocide); 2) because of intensive language
contact. One group in contact with a second group disappears because it makes the shift to the other group’s
language.
Language interference > It is a term to refer to any negative influence (lexical, syntactic, phonological, etc.) that one
language exerts over the other. Interference usually hinders the learning process and causes a problem to the
language learner.
Language maintenance > It refers to the preservation by a speech community of its native language from generation
to generation. It includes maintaining the languages of cultural minority groups through family practices, rituals,
concerted educational endeavours with society at large, etc. The language can change but only by small degrees and
owing to internal developments and/or contact with other languages.
Language mixing > It occurs when the interlocutors change from one language to the other in the course of a single
conversation and more precisely when switching back and forth occurs within a clause.
Language shift > It is a process whereby contact with other languages can lead to the partial or total abandonment
of a group’s native language in favour of another with a higher status. In some cases, the minority language may
preserve its L1 for certain functions, while acquiring the dominant language for other uses.
Lingua franca > It is a language used as a means of communication between populations speaking vernaculars that
are not mutually intelligible. English is at present the most widespread lingua franca.
Loanword > It is a word adopted from a donor language and incorporated into a recipient language without being
translated. It shows how languages influence one another due to the contact between them.
Mixed language > It is a language that arises through the fusion of two source languages, normally in situations of
bilingualism. It often retains the inflectional complexities of both parent languages. In a mixed language, both source
languages are clearly identifiable, and it may mark the appearance of a new ethnic or cultural group.
Pidgin > It is an auxiliary language which doesn't have native speakers. It has a reduced structure and lexicon. It is
employed for communicative purposes between two groups of people, one usually superior to the other in
economic and political power. It is used as a lingua franca in situations such as trade, or where both groups speak
languages different from the language of the country in which they reside.
Tok Pisin > It is one of the three official languages of Papua New Guinea, and the most widely used language in
that country. It is sometimes used as a pidgin and sometimes as a creole, and shows clear influences from English
in terms of borrowings, subordination patterns, or plural forms.