Censuses are faced with all of the policy problems that plague anything
that deals with language. It may be instructive to think about
censuses and surveys in terms of who?, what?, why?, when?, where?
and how?
What? Censuses are concerned with enumeration, and must
be able to work with data that can be processed according to pre-existing
definitions of the categories asked about. Census takers (who?) must
therefore decide in advance what is a language and what is not,
and ask questions about language that can be easily enumerated.
Who? The actors in a census are
The census takers, representing a jurisdiction of some sort:
a state, nation, school district, territory,
whatever.
The subjects, i.e. the people being enumerated, questioned,
surveyed; in this case they are "speakers" of (a) language
(s) and are asked questions about their ability in, knowledge of,
proficiency in, loyalty to, literacy in (...) one or more
languages.
The languages in question; this immediately raises
problems with the age-old lack of clarity about what is a language
and what is not, and what other categories or subgroupings might be
used.
Most polities ask questions that fulfill pre-existing expectations,
and will not ask questions that result in difficulties of enumeration, or
challenge pre-existing notions of what is language, what is the
national/official language, etc.
When? Surveys are undertaken periodically, perhaps
decennially (US, India, etc.) perhaps only occasionally, or when
confronted with some sort of "problem". Some polities will attempt a
linguistic census as a precursor to some policy decision; India in the
1910's and 20's (Grierson) attempted to classify languages so as to better
deal with census issues. Other polities (see below) are looking for some
kind of change in linguistic behavior, or perceive
existence/persistence of some language/variety as a threat to national
security, etc.
Therefore most polities ask questions about the approved
languages of the polity, the national or official
(however defined) language, the mother tongue and are not
interested in variances from these categories.
Most polities assume people speak the languages their
parents speak, and are not interested in variance from tradition, i.e.
they are not interested in language shift unless shift is an overt
or convert goal of the polity (get rid of those pesky X speakers).
Most polities want either/or (yes/no) answers about language,
and are not interested in bilingualism or multilingualism, or answers that
complicate simple either/or questions.
Where? Polities tend to prefer that if there are linguistic
minorities, that their distribution be territorial , and therefore
will not ask questions about the distribution of languages outside
traditional territories, for fear that the very question of portability of
language (personal right) might induce somebody to demand that
right. Polities prefer that linguistic territoriality be confined
to marginal or peripheral areas of the polity, i.e. not in the
center of cities or centers of power.
How? What is the methodology and procedure of a language
survey, and is it ever done without ulterior motives (covert policy
assumptions) or without predetermined goals in mind?
Why do some multilingual polities not ask questions
about language?
Polities that do not consider themselves multilingual, or do not have
multilingualism as a goal do not usually ask questions about language:
best example: France.
This does not preclude France from
occasionally doing statistical
surveys (performed by the I.N.S.E.E.) in a
particular region, such as the survey in the Alsace/Mosel region in the
late 50's, when it was finally determined that only very aged people
were bilingual, and all younger people spoke French.
The US has not often asked linguistic questions in its
decennial census, but in 1916 and in 1940, the Census Bureau did surveys
(off-year censuses in the case of 1916) on the "mother-tongue of the white
population." Could these special census have had anything to do with the
fact that in the following year (1917 and 1941) the US
entered wars? The US Census began asking questions again in a sporadic
manner in the late 20th century; the 2000 census also paid some
attention to these issues, as
Crawford's article here points out.
Avowedly multilingual polities (e.g. Switzerland, Belgium,
(former) USSR, India, etc.) may ask questions about language, but almost
always in expectation of fulfillment of previously defined expectations.
Strict territoriality, or an attempt to determine actual linguistic
boundaries (conceived of as either/or) are the underlying assumption in
polities with territorial principles.
(Exception: in early Soviet period, surveys were performed in vacuo, in
an attempt to elicit information for the formulation of language policy;
subsequently, new questions not asked.)
Malaysia is interested in only one linguistic answer: what progress
are non-Malays making in learning Malay? No other answers (e.g. Tamil
speakers (non-Malays) have ceased to be speakers of their former mother
tongue (Tamil) and have become English speakers) are of interest.
Wars, immigration, and other kinds of disruption or conflict
may result in changes in census policy ; after World War II,
Germans in Soviet Union were de-categorized, removed from their
Autonomous Regions, shipped to Central Asia (as were some other suspect
nationalities, such as Balkarians etc.) Treatment of Jews also changed;
Karelians were downgraded from Republic status.
Laws of unintended consequences: India redrew state borders
in the mid-50's, based on the census of 1951, but this led to exacerbated
demands for more autonomy and even secession, rather than calm and order.
The 1961 census accepted any declaration whatsoever of language that
anyone wished to make, but resulted in many "spurious" declarations that
had then to be discarded or changed. In the cases of supposed Hindi
dialects, "spurious" dialects were merged with Hindi data; but other
minor languages were not merged with other languages. in the 1971 census
this latitude was not allowed, and declarations by small groups were
simply ignored.
School and Educational Surveys.
Here are some reports of educational surveys
chosen at random from ERIC/CD-ROM reports. Note what the "subject
major" and "subject minor" of the report is concerned with: not just "who
speaks what to whom, where, what and when," but also limited English
proficiency and other issues.
Linguistic Censuses. These are censuses of an official or
semi-official nature, or done by demographers and/or geographers.
Grierson's Linguistic Census of India
Census of India, 1961, Vol. 1, Part II-c(ii) Language
Tables
Here is a page from the 1990 Singapore
Census, showing predominant household language of various linguistic groups.
This page is a form used in the Soviet Census of 1989
(the very last Soviet census), with the original sheet (2nd page) preceded by a sheet with
translated notes for various questions.
Breton's Atlas Geographique des Langues et des Ethnies
de l'Inde et du suncontinent
Sociolinguistic Surveys
These are surveys
conducted independently of "official" studies by census bureaus or
official governmental organs; they focus more on sophisticated kinds of
sociolinguistic competence not usually enumerated in censuses.
The 1790 survey that resulted from the Cahier des
Doléances during the French Revolution, which led the Abbé
Grégoire to propose language legislation that would ensure the
francisation of regions where French was not spoken.
