Zooming in on Longitudinal Variation

Studies of language change traditionally rely on two kinds of methods. The first and oldest involves
systematic comparison of evidence from different historical periods of a language. Typically, the evidence is
drawn from written texts, which are usually written by different authors at widely separated periods. Such
studies face the challenge of arguing that the texts used are comparable enough to justify their comparison,
and to license the inference from observed variation to change. But numerous contextual factors, such as text
genres, author identities, regional dialect differences, etc. confound such comparisons (Herring, et al. 2001).
The second type of study looks at synchronic variation in a speech community and compares the age
distribution of linguistic features among members of the community to draw inferences about earlier and later
historical periods. Such studies must assume that language variation is relatively stable within any given
individual, or there would be no reason to believe that an individual’s current linguistic behavior accurately
reflects an earlier time period.
Consequently there are large gaps in our understanding of language change. To what extent is
intra-individual variation implicated in language change? And to what extent does intra-individual change
accompany historical change? Some recent studies of language variation and change partially address these
issues by combining community studies with longitudinal data collection (e.g., Yoneda 1993), others develop
large-scale statistical models of multiple authors over time (e.g., Burrows 1992, Finegan and Biber 1989), and
even others employ corpora of spoken data from radio (Van de Velde et al. 1998) and cinema (Elliott 2000)
controlling for effects of medium and genre. Yet, the field still lacks studies comparing continuous
longitudinal variation to individual variation.
Two recent developments, both linked to changes in computer technology, facilitate a new kind of
study that can begin to answer these questions. The first is the ability of modern computers to store and
analyze large amounts of language data. Corpus linguistics is emerging as the preferred paradigm for studying
language variation and change: not only can the computer assist in locating data items of interest in large
compendia of computerized data, but the ability to handle large amounts of data helps us control for more
contextual factors and arrive at more reliable statistical analyses. The second development is the widespread
adoption of the Internet as a communications medium. People with a broad range of personal and professional
backgrounds use the Internet for primary communication in a wide range of purposes. Much of the material is
archived so that continuous records lasting more than ten years are now becoming available. Through careful
selection of archived materials, it is possible to construct corpora of continuous Internet communications that
control for contextual factors such as genre as well as individual usage. Furthermore, the Internet is believed
by some to be an active force for language change within English, by promoting certain forms of abbreviation
and ellipsis, increased informality, decreased politeness, and the like (Baron 1984, 2000; Crystal 2001).
Currently available archives of Internet communication hold the key to understanding whether these things
are in fact true. In addition, the ability to view continuous records of language use offers an unprecedented
opportunity to observe the relation between individual variation and language change.
In this paper, I examine variation among 21 different linguistic variables over the complete 11-year
history of MsgGroup, and early ARPANET/Internet email discussion group established in 1975 to discuss the
protocols of electronic mail. The linguistic variables were selected from the 67 features in Biber (1988), so
58
that candidate data items could be readily identified automatically, and so that the variation in the electronic
mail corpus could be readily compared with a comprehensive study of English variation. The variables were
studied both in aggregate, using factor analysis, and individually, using variationist models. Messages from
each of the 329 participants in MsgGroup were tracked over the entire 11-year span, so that individual
longitudinal trends could be observed and compared to the overall trends.
What emerges from this study is a very complex picture of language variation which points not to
language change, but rather individual adaptations based on immediate and local communicative
circumstances. In the factor analysis, three fairly robust factors emerge: a factor of general elaboration, one of
lexical and syntactic complexity, and one of person reference and epistemic certainty. None of these factors
correlates strongly with historical time, and the variance on all factors remains high throughout the 11-year
observed span. In other words, the main shared patterns of variation in the corpus do not turn out to be
patterns of change over time. Individually, however, some of the linguistic variables, such as first and second
person pronouns, are significantly correlated with historical time, and when patterns of individual participants
are analyzed, yet other significant trends are observed. In many cases, such as the sharply decreased use of
contractions over time by one senior female system administrator, these trends run counter to the trends of
most other group members. Sometimes these trends are traceable to specific events, such as a particularly
active and divisive debate that occurred in on MsgGroup in 1979 concerning the finger protocol. Individuals
appear to adapt their use of linguistic variables in response to changes in status and social affinity that arise
from these events.
Zooming in on longitudinal language variation thus reveals a complex situation in which individual
variation and historical change do not necessarily recapitulate one another, as suggested by prior models.
Instead, individual patterns of variation appear to respond to complex local and political concerns resulting in
broad intra-group variance when viewed continuously over time. These findings thus challenge us to propose
theories of language change that successfully relate such complex continuous variation to long-term historical
trends.
REFERENCES
Baron, Naomi. 1984. “Computer-mediated communication as a force in language change”. Visible Language
18: 118–141.
Baron, Naomi. 2000. Alphabet to email: How Written English Evolved and Where it’s Heading. London:
Routledge.
Biber, Douglas. 1988. Variation Across Speech and Writing. Cambridge: Cambridge University Press.
Biber, Douglas and Edward Finegan. 1989. “Drift and the evolution of English style: a history of three
genres”. Language 65: 487–517.
Burrows, J. F. 1992. “Computers and the study of literature”. In C. S. Butler, ed., Computers and Written
Texts, 167–204. Oxford, UK: Blackwell.
Crystal, David. 2001. Language and the Internet. Cambridge: Cambridge University Press.
Elliott, Nancy. 2000. A Sociolinguistic Study of Rhoticity in American Film Speech from the 1930s to the
1970s. Doctoral dissertation, Indiana University.
Herring, Susan, Lene Schosler, and Pieter von Rainen. 2000. Textual Parameters in Older Languages.
Amsterdam: Benjamins.
Van de Velde, Hans, M. Gerritsen & Roeland van Hout. 1996. “The Devoicing of Fricatives in Standard
Dutch. A real-time study based on radio recordings”. Language Variation and Change 8: 149–175
Yoneda, Masato. 1993. “Survey of standardization in Tsuruoka City, Japan: Comparison of results from three
surveys at twenty-year intervals”. Paper presented at Methods VIII: International Conference on
Dialectology. University of Victoria, British Columbia.