Fourth Language and Politics Symposium
Queen's University of Belfast: 18th - 19th September 2003
Celtic Languages in the 2001 Census: how population censuses bury Celtic speakers
Kenneth MacKinnon
Introduction
Questions on Celtic languages have featured on U.K. population censuses since 1851 for Irish, 1881 for Gaelic and 1891 for Welsh. From 1921 age-analysis of speakers has been presented, but with the partition of Ireland at that time, no information on Irish speakers in Northern Ireland was sought from that census nor until 1991. Since 1971 data on reading and writing abilities have been sought, and in 2001 for the first time, ability to understand these languages. The census question has been limited to the countries in question and the census authorities have resisted requests to seek information on abilities in Celtic languages throughout the U.K.
Disparities in the language question over this period (seeking response on whether the person can speak Gaelic or Irish, but does the person speak Welsh) were resolved in 2001 with the adoption of can as the operative question in all cases. Actual numbers of Welsh speakers may thus have been tacitly depressed in previous censuses.
This paper reviews the resources of the three national census authorities with regard to results published to date (September 2003) on the distribution of abilities in the three Celtic languages: Welsh, Gaelic and Irish in Northern Ireland by:-
• Office for National Statistics for England and Wales (ONS);
• General Register Office for Scotland (GROS); and
• the Northern Ireland Statistics and Research Agency (NISRA)
These results relating to Celtic languages are presented in-
• univariate tables, which detail totals of speakers and other language skills;
• standard tables, which detail speakers and other combinations of abilities by age and sex, together with a further factor (such as country of birth, general health and limiting long-term illness, etc.);
• theme tables which detail age and sex breakdown of speakers or persons with any language abilities aggregated either in terms of numbers and/or percentages in terms of a range of social, economic and demographic factors; and
• key statistics tables, which are generally outline tables of specific factors (e.g. aggregate of language abilities) as percentage of total populations of areas.
Although there has been some measure of common practice between these authorities, there is nevertheless a considerable degree of disparity between them in the way in which they present language data, such that it is often very difficult to use them to ascertain actual numbers of speakers, and to compare these languages in their respective societies in meaningful ways. For example age-groups do not always ‘map' or are not always consistent between tables of a given national census authority, nor between comparable tables of different authorities. In some cases, detailed below, numbers of actual speakers are aggregated into an ‘other combinations' category, which makes totaling actual numbers of speakers impossible. Consistent practice with previous censuses has not always been followed, which makes trend analysis particularly difficult. The tables for the 2001 census published or announced by September 2003 are reviewed below.
Welsh Language in the 2001 Census
Data on Welsh language abilities are available in the following published tables:-
• Univariate table UV84 presents totals of speaker aged 3+ (total for Wales = 582,368), and the three other language abilities separately for Wales and local authority areas. There is of course an unknown overlap between them, and it would be helpful for the 16 combinations of the four language abilities to have been detailed.
• Standard table S133 details sex and age by country of birth and knowledge of Welsh, with 12 age-categories (3-4, 5-9, 10-14, 15, 16-19, 20-24, 25-34, 35-49, 50-59, 60-64, 65-74, 75+), and 6 ability categories. Welsh speakers in Wales total 575,640, and it is clear that a total of 6,728 Welsh speakers are contained in the ‘Other Combinations' category and are thus inaccessible or unanalysable. (Results for 1991 census were presented in LBS Table 67W in age-categories 3-4, 5-10, 11-15, 16-17, 18-19 thence quinquennially to 85+) Finesse in comparison of age-structures is thereby reduced.
• Standard table S137 details language abilities, sex age, limiting long term illness and general health. There are four age categories ( 3-15, 16-44, 45-64 and 65+). There are five language abilities categories including an ‘Other Combinations' category which contains an undisclosed 6,728 Welsh speakers nationally. The national total of speakers adds up to 575,640.
• Standard table S143 presents data on children speaking Welsh in Welsh-speaking households. Similar tables for Gaelic and Irish would have been an excellent idea too. In 1991 Topic monitors for Welsh and Gaelic did in fact do so – and included the cases of Welsh and Gaelic speaking ability amongst children whose parents did not speak the language.
• Theme table T15 aggregates all Welsh language abilities into a single category, totaling 797,717 for Wales as a whole. Analysis is in terms of 7 age-categories (3-4, 5 –15, 16-19, 20-44, 45-64, 65-74, 75+). Cross-tabulation is presented in terms of 17 demographic, social and economic factors: sex, accommodation, occupancy, amenities, central heating, car or van, living arrangements, family type, economic activity, country of birth, ethnic group, general health, long-term illness, qualifications, industry, and socioeconomic class. These details enable a way-of-life / quality of life analysis to be undertaken.
• Theme table T39 more usefully cross-tabulates all actual Welsh speakers aged 3+ ( = 582,368 for Wales as a whole) in terms of the same categories as T15.
• Key statistics table KS25 details numbers and percentages of Welsh speakers (=575,640 for Wales as a whole) for Wales and local authority areas. There are 6 skills categories which again contain 6,728 Welsh speakers in the ‘Other Combinations' category.
• A ranking table of local authorities in terms of percentage of population with one or more Welsh language skills was published at an early stage.
It would have been extremely helpful if the definition of ‘Welsh speaker' adopted for the univariate tables were used throughout. This would enable analysis of Welsh speakers to include all instead of the 98.8% of such which feature in tables S133, S137 and KS25. Table T39, which does include all Welsh speakers, is limited to 7 age-categories only.
Gaelic Language in the 2001 census
Data on Gaelic language is available in the following published tables:-
• Univariate table UV12 presents totals of all persons with 9 combinations of language abilities. The ‘Other combinations of skills in Gaelic' category is in fact the total of persons who can speak and write but cannot read Gaelic. It is thus possible to derive totals of all Gaelic speakers from this table. The total is 58,969 for Scotland as a whole. It must be emphasized that this total is for all people and includes those aged 0-2 years. Data is available for Scotland as a whole and all local authority areas.
• Standard table S206 details Gaelic speaking abilities by age, sex and whether or not born in Scotland . There are 7 ability categories from which it is possible to derive totals of actual speakers as such, persons without speaking skills but ability to read and/or write Gaelic, and persons with understanding ability only. There are 20 age-categories ( 0-2, 3-4,5-11,12-15,16-19, thence quinquennially to 90+). Similar age-categorisation would have been desirable in the case of Celtic languages in the other national censuses.
• Theme table T27 presents data on Gaelic speakers as such in comparable terms to the ONS tables T15 for all Welsh language skills and T39 for all Welsh speakers. This will enable some valuable comparative analysis to be undertaken on Welsh and Gaelic speakers. The age-ranges are useful but could be more detailed. This cannot be attempted for Irish in Northern Ireland as things stand, since the theme table categories for Irish in T32 do not match those for Welsh and Gaelic.
• The Key Statistics table KS06 presents a total of all persons with Gaelic language skills as a single grouping and whether or not born in Scotland as percentages of total populations of Scotland and local authority areas.
• A Gaelic language table features in the Registrar General's Report to the Scottish Parliament, Table RG10. This presents data on totals of persons able to speak Gaelic, speak, read or write Gaelic in 1991 and 2001, and all persons able to speak, read, write or understand Gaelic in 2001 for Scotland and local authority areas. The table totals 58,652 actual speakers, some fewer than the UV12 total of 58,969. The category of ‘speaks, and writes but does not read Gaelic' is obviously (and inexplicably) omitted.
• There is also a table on aggregates of all Gaelic abilities, and whether born in Scotland as a percentage and as numbers for about 630 major towns and settlements.
As published to date, the GROS data on Gaelic is more finely tuned and more consistent than corresponding Welsh data. There is a specific table for Welsh in the family which would be valuable too for Gaelic in line with 1991.
Irish Language in Northern Ireland
Initial data on Irish language became available in 2002. The published tables were not comparable in detail with the five tables published in the 1991 Irish Language Report. These tables were in terms of 8 ability categories from which actual numbers of speakers could be totalled. The tables comprised:-
• Table 1: Age table at province and district levels (single years of age to 24 years, thence quinary age-groups to 95+); and at provincial level only:
• Table 2: Religion and sex by knowledge of Irish; and for the 16+ population:
• Table 3: Qualifications by knowledge of Irish;
• Table 4: Economic activity, by sex and knowledge of Irish; and
• Table 5: Social class by sex and knowledge of Irish.
Continuity with these tables would have been useful. What is currently available comprises:-
• Univariate table UV14 detailing the full 16 categories of Irish language abilities. This is available at provincial and district levels – and will be available down to neighbourhood (output area) levels in terms of a simplified 9 category breakdown, comparable to the Gaelic Table UV12. The total of actual Irish speakers which can be derived from UV14 is 115,731 (compared with 131,974 in 1991).
• Standard table S372 details sex and age by knowledge of Irish. The 7 Irish language skills categories include a ‘some knowledge of Irish' category, and an ‘Other Combinations' category, which totals 24,167 for the province, and would seem to contain 15,280 actual speakers. There are 7 age-categories: 3-11, 12 – 15, 16-24, 25-30, 40-50, 60-74, 75+. These are far less detailed than the 1991 Table 1, and only allow for the most ‘broad brush' trends analysis. The lack of single years of age for Irish speakers under 25 means that any assessment of success or effects of Irish-medium school level education will be extremely limited, with very little possibility of finesse.
• Standard table S373 details sex and age by knowledge of Irish and country of birth. Language ability is a simple dichotomy: any knowledge of Irish or none. There are 4 country of birth categories: Northern Ireland , Republic of Ireland , Ireland part unspecified, and elsewhere. The 7 age categories are as in S372.
• Standard table S374 details sex and highest level of qualifications by knowledge of Irish for persons aged 16-74 only, as a single group. There are 5 qualification levels and age categories as in S372, which does not enable actual numbers of speakers to be inferred
• Standard table S375 details sex and age by knowledge of Irish and ‘Community Background' (Religion or Religion brought up in). Language abilities are a simple dichotomy of any knowledge of Irish or none, with 7 age-categories as in S372. There are 4 Community Background categories: Catholic, Protestant including other Christian and related, Other religions and philosophies, and none. This categorization is supplemented by a more specific range in:
• Standard Table S375A which details sex and age by knowledge of Irish and Religion.. Ability and age categories are as in S375, but the religious affinities are detailed as: Catholic, Presbyterian, Church of Ireland, Methodist, Other Christian and related, Other religions and philosophies, No religion or not stated. It is not possible to infer actual speakers from S375 and S375A – although the broad age-structures may enable some analysis of differential language-transmission rates to be attempted between the various community and religious groups..
• Theme table T32 on Irish language presents a simple dichotomy of any knowledge of Irish or none – and not in terms of actual speakers as do comparable theme tables for Welsh T29, and Gaelic T27. The common age-categorisation of these tables is not followed for Irish, which is: 3-11, 12-15, 16-24, 25-39, 40-59, 60-74, 75+.. The only category which matches the Welsh and Gaelic theme tables is the 75+ group, so it is not possible to attempt any comparative analysis of social, economic, way-of-life or quality-of-life characteristics of the three UK Celtic language groups. There are 17 social and economic factors, which include ‘Community Background'. There is no Ethnic Group category – as this was not asked in Northern Ireland .
• Key statistics table KS24 details knowledge of Irish in terms of numbers and percentages, by province, district, education and health board areas, and parliamentary constituencies. There are 7 language ability categories as in tables S372 and S374, from which it is not possible to infer actual numbers of speakers.
There is a clear need for tables to total actual numbers of speakers as such, as well as understanders, readers and writers of Irish. Age tables of Irish speakers for single years of age to 24 years (and quinary age-groups thereafter) for the four basic language skills at least, at Education Board level would be invaluable in order to monitor the effectiveness of Irish-medium school-level education. An Irish language theme table in terms of actual speakers as for Welsh and Gaelic, and in terms of the actual age-categorisation .used for those languages as a standard UK practice would be highly desirable. So too would be a table on Irish in the family and household structures on lines of Welsh language table S143.
Applications and difficulties: age-analysis
The population census represents an invaluable resource for studying the incidence of the language abilities in these three cases, and the social and economic characteristics of these three language-groups. It is a great pity that the shortcomings identified above make much of this work difficult.
For example, comparison of Irish speakers in 1991 and 2001 is strictly speaking impossible to illustrate. The 37 age-categories in 1991 can largely be mapped to the 7 in 2001 but there is considerable loss of detail. Single year of age data to age 24 for 1991 do however make the production of 3-11 and 12 – 15 age-groups possible to compare with these groups in 2001. However from age-related data so far published it is not possible to produce actual numbers of speakers as such for 2001. Moreover, the age-categories are irregular, covering cohorts of 9, 4, 9, 15, 20, 15 years in the under-75 years range. This grossly distorts visual impression, thus making the understanding of language data, and language planning both very difficult.
Similar difficulties affect the comparison of Welsh speakers between 1991 and 2001. Again, the best source (Table S133) omits 6,728 Welsh speakers. The table has 11 age categories under 75 years, with cohort sizes of : 2, 4, 5, 1, 4, 5, 10, 15, 10, 5, and 10 years.. Illustrating these data as they stand again produces gross visual distortion of the impression of age-distribution of speakers. It is still unsatisfactory and cannot really be compared with the age-distribution for Irish in Northern Ireland . The lack of single year of age data for under 24s means that the age-groups of 3-4, 5-10, 11-15 years in 1991 cannot very satisfactorily be mapped onto the age-groups of 3-4, 5-9, 10-14, and 15 years in 2001.
It is possible to produce a detailed comparison of age-distribution of Gaelic speakers between the 1991 and 2001 censuses from Gaelic language data in the 1991 Census tables LBS 67S and the 2001 table S206. These age data are much more satisfactory than corresponding Welsh or Irish data, in that the age groups of 3-4, 5-11, 12-15 in the 1991 table LBS 67S match those in the 2001 table S206. It is a pity that data so far available for Irish in Northern Ireland or Welsh in Wales do not enable a comparable picture to be presented. It would obviously enable a better evaluation to be made of the relative effectiveness of Welsh-, Gaelic- and Irish-medium educational practice, and the comparative viability over age of these three language groups.
The inclusive character of all Gaelic speakers in the data on Gaelic language in Table S206 also enables the incidence of language abilities to be graphically illustrated. Actual speakers can be totalled, as can non-speaker readers and/or writers, and understanders. The Gaelic data are available in 20 age-categories, compared with 12 for Welsh and 7 for Irish. A far more detailed and discriminating picture can be presented for Gaelic language abilities than can be done either for Welsh or for Irish. It would be valuable to be able to undertake this level of analysis for these two languages also.
There has been a very marked effect of increases of Gaelic speakers in the primary education age-group. The fact that Gaelic-medium education has not been carried on into the secondary stage is again very obvious. Weaknesses in the 16-19 age-group, which result from this, also show up very conspicuously.
Reversing Language-Shift
Census data also enable measures of language maintenance, language vitality and reversing language-shift to be derived. These are invaluable as a means of comparing the situation in different areas – and could even be a means of comparing the fortunes of the different language groups in their respective countries.
United Kingdom census data can be used to operationalise Joshua Fishman's (1991, 2001) concept of reversing language-shift (RLS) by comparing mean values of language ability data for the parental and childhood generations. A measure entitled Intergenerational Gain / Loss has been derived from the proportion of minority language speakers of the parental age –range in a particular population, compared with a similar calculation of the proportion of minority language speakers within the age-group of childhood. The salience of each age-group within a specific population is thus compared. Positive values indicate language gain and negative values indicate loss. This assumes some stability of populations, which today is generally very far from being the case. The measure does however indicate salience of the language-group in local populations. A more reliable measure is Intergenerational Ratio which compares mean numbers as a single year-of-age in each age-group as a ratio of the numbers of speakers within the age-group of childhood to those within the parental generation. Values above unity indicate outright gain, and values below unity indicate actual loss, irrespective of other societal factors..
It is possible to plot both values graphically, thus indicating trends. It has been possible to compare corresponding age-groups for language-groups as a whole and between the two censuses, namely 5-15 and 20-44s. Again, the 2001 values for Irish speakers aggregates ‘speakers' as such with ‘other combinations' and is therefore not accurate. It is a great pity that the Irish language theme table T32 did not enumerate totals of actual speakers as did its Welsh language counterpart T39. Moreover the available age-tables (e.g. S372) do not enable age-groups of Irish speakers to be matched with the other Celtic languages. So the illustration can be impressionistic only – rather than strictly comparable or accurate.
The difficulties here arise from the non-comparability of age-groups owing to discrepancies of age-groups between the Gaelic table S206 which was available for council areas, and the most closely comparabe Welsh table which was similarly available at the time, S137. When – or if - T39 becomes available at local authority levels, this could be rectified.
This exercise also enabled mapping to be attempted, and distribution maps for Intergenerational Gain and Intergenerational Ratio for Welsh in 2001 are shown as Figures below. The considerable gains in numbers of Welsh speakers owing to educational policies stand out very clearly – especially in anglicized southeast and south Wales . Similar mapping has been undertaken for Gaelic but until more reliable figures for Irish speakers by age and local areas are forthcoming, this is not really possible for Irish in Northern Ireland .
Recommendations
• Include all Welsh speakers in Welsh language tables S133, S137 and KS25
• Similarly include all Gaelic speakers in Table RG10.
• In Northern Ireland present actual numbers of Irish speakers in all Irish language tables. In particular:
• In Irish-language theme table T32 to show all actual speakers, and use same age-categories as Gaelic T32 and Welsh T39 theme tables.
• Standardise all Welsh, Gaelic and Irish theme table age-categories. The following categorisation is recommended: 3-4 Pre-school, 5-11 Primary, 12-15 Secondary, 16-24 FE/HE/Young adult, 25-34 Younger parents, 35-44 Older parents, 45-54 Early Middle-age, 55-64 Older Middle-age, 65-74 semi/early retired, 75+ older. These categories more adequately reflect life stages and are more regular or even than the existing categories.
• Age-breakdowns for standard tables: 5 year age-groups – or where impracticable as for theme tables (as recommended above.) Use GROS standard table S206 as a model.
• In Northern Ireland an age-table for Irish speakers by single years of age for under-25s as in 1991 Table 1 would enable effects of Irish-medium schooling to be effectively assessed. Similar single year-of-age tables for speakers under-25 for Welsh and Gaelic would do likewise. Standardise on best practice.
• For language-ability categories in univariate tables, present the full 16 categories in each case, i.e. standardise on the recently-announced Irish-language model for higher levels of census geography (NI UV14). A nine-category model as for GROS UV12 for lower levels would be useful where the 16 categories are impracticable.
• In other tables with language-ability categories enable all speakers to be identified. At the minimum, enable all speakers; readers, writers and understanders to be totalled, and similarly also all readers, writers and understanders without speaking ability.
• Produce tables on Gaelic and Irish in family / household structures on similar lines to Welsh data – and in line with 1991 tables in Topic Monitors.
• Produce special topic reports / monitors on the three language-groups in comparable methodology, incorporating the above recommendations.
References and sources
Fishman, Joshua A. (1991) Reversing Language Shift, Clevedon: Multilingual matters ISBN 1-85359-121-1 ( pbk.)
Fishman, Joshua A. (Ed.) (2001) Can Threatened Languages be Saved? Clevedon: Multilingual matters ISBN 1-85359-492-X ( pbk ).
General Register Office for Scotland , Edinburgh : 1991 Census Table LBS 67S; 2001 Census tables: KS06, S206, T27, UV12, Registrar General's Report to the Scottish Parliament, Table RG10.
Northern Ireland Statistical Research Agency, Belfast , 2001 Census tables KS24, S372, S373, S374, S375, S375A T32, UV14.
Office for National Statistics for England and Wales , Titchfield, Fareham , Hants., 1991 Census table: LBS 67W; 2001 Census tables: KS25, S133, S137, S143, T15, T39, UV 84
Registrar General Northern Ireland (1993) The Northern Ireland Census 1991: Irish Language Report, Belfast : Her Majesty's Stationery Office (No ISBN)
Acknowledgements
Considerable thanks are due to the helpfulness of staff at the various census offices, their readiness to respond to requests for information, promptness in sending available material electronically, and also readiness in receiving suggestions for development of products and format of tables. In particular I should like to thank Ruth Fulton at NISRA, Penny Wilkinson at ONS, Frank Thomas at GROS, and Trudy Culshaw at IOM Census Office..
________________________________________________________________________
Kenneth MacKinnon is Visiting Professor and Emeritus Reader in the Sociology of Language at the University of Hertfordshire , Honorary Fellow in Celtic at the University of Edinburgh , and Associate Lecturer in Social Sciences, Education and Language Studies at the Open University.
Contact details : SGRÙD Research, Ivy Cottage, Ferintosh, The Black Isle, by Dingwall, Ross-shire IV7 8HX Tel / Fax: 01349-863460 E-mail: kenmackinnon@enterprise.net Website: www.sgrud.org.uk
________________________________________________________________________