MOTHER TONGUE WELSH SPEAKERS AND WELSH SPEAKING COUPLES

MOTHER TONGUE WELSH SPEAKERS AND WELSH SPEAKING COUPLES Howell M Jones, Y Swyddfa Gymreig / Welsh Office Views expressed in this paper are the personal views of the author. Abstract The 1992 Welsh Social Survey was the largest survey ever, with the exception of the Census of Population, to enquire into the Welsh language abilities of the populace. Whereas the Census asks only of an individual's current use of Welsh the 1992 Welsh Social Survey asked also about the language used at home as a child. This permits analysis of mother tongue speakers of Welsh, a mother tongue speaker being defined here as someone who as a child spoke more Welsh than English at home. The paper presents results pertaining to the characteristics of mother tongue speakers in terms of current ability in Welsh and the language of their parents. The percentage of mother tongue speakers by age of speaker is graphed and the validity of the results assessed by comparison with an alternative source of statistics. The prevalence of Welsh speaking couples, a determining factor for the continuance of mother tongue speakers, is examined. Introduction The future of the Welsh language as a natural medium of communication for a significant proportion of the population of Wales clearly depends on the degree to which the language is transmitted to each new generation. Of the means by which a language can be transmitted, it is generally accepted that language reproduction within the family is the most important. This issue of language reproduction within families has been examined in Wales from both sociological and psychological viewpoints. Quantitative studies have largely been dependent on the Census as a data source. Although broadly satisfactory for analyses concerning the spatial distribution of Welsh speakers, and long term trends, the limited range of questions asked about the language in the Census has not permitted quantification of many aspects of inter-generational language transmission. The 1992 Welsh Social Survey has provided a means by which such quantification may be achieved. Most mother tongue speakers come from families where both parents speak Welsh. The percentage of the population who are mother tongue speakers will depend in the future both on the intergenerational transmission rate and the percentage of couples who are Welsh speaking. The paper presents results which shows how both the percentage of mother tongue speakers within the population, and the percentage of Welsh speaking couples of all couples, has varied over time. The relationship between the two is also examined. In respect of children of primary school age, an alternative set of estimates of the proportion who speak Welsh as a mother tongue exists, derived from primary school statistics. The two sets of estimates are compared and generally the comparison is encouraging. This validation provides support for the credibility of the series derived from the social survey. A technical annex describes the method of imputation used to enhance the initial partial data set so that the mother tongue variable was available for the entire data set. Background to the 1992 Welsh Social Survey The 1992 Welsh Social Survey was carried out in the last quarter of 1992 and inquired into housing, education and training as well as Welsh. With a sample exceeding 1 per cent of the population, selected randomly using a stratified design, it produced estimates of the numbers speaking Welsh which were close to those produced by the 1991 Census. However, the survey asked over 30 language related questions which permits a range of analyses which can not be obtained from the Census. This paper examines mother tongue speakers of Welsh, their current ability in Welsh and their parental linguistic background. A further original analysis looks at the numbers of households containing a couple where both partners speak Welsh, by the average age of the couple. The percentage Crown Copyright 1995 1

representation of Welsh speaking couples is also compared with a simple model to examine the phenomenon of language group endogamy. It is estimated that in 1992 there were 347.6 thousand mother tongue speakers (12.6 per cent of the total population): 94 per cent of these still assessed themselves as Welsh speakers. Of current non- Welsh speakers, 1 per cent spoke Welsh as their mother tongue Of the total number of current Welsh speakers, 55.3 per cent (326.6 thousand) were mother tongue speakers of Welsh. However, the predominance of mother tongue speakers increases with claimed degree of use, so that 58.1 per cent of occasional fluent speakers were mother tongue speakers, whereas 93.2 per cent of those fluent speakers who claimed to speak Welsh all or most of the time were mother tongue speakers. This preponderance of mother tongue speakers amongst intensive users of the language is the primary reason the number of mother tongue speakers may be considered significant for the future of the language. (There is also the possibility that mother tongue speakers should be considered as key to the future of the language in terms of syntax and lexical content etc. though recent evidence in Wales is mixed e.g. Mathias). 1992 Welsh Social Survey questionnaire The questionnaire used to ask about Welsh was constructed with 3 main sections. The first section included two questions about whether the mother and father spoke Welsh while the last question in the section asked for self-categorisation according to ability in, and use of, Welsh. Routing to the subsequent two sections depended on the response to this question. All claiming to be fluent in Welsh were to jump the second section which dealt with understanding and intentions towards learning Welsh. The third section started with the question "As a child at home did you speak more Welsh than English?". Affirmative responders to this question are referred to in this paper as mother tongue speakers of Welsh. It was recognised at the design stage that the filtering process could lead to significant numbers of people inappropriately failing to answer the third section but the priority was to ensure maximum response to the second section of the questionnaire. In the event, significant numbers did fail to answer the third section and so, in order that a full range of analyses of mother tongue speakers could be undertaken, it was necessary to impute responses for those who had not answered the question "As a child at home did you speak more Welsh than English?". The method by which this was done is described in the Technical Annex to this paper. The imputation was achieved by making use of the correlation between speaking Welsh as a mother tongue and two other facets of the individual: current ability in the language, and the language spoken by the individual's parents. Results are presented below which demonstrate these relationships. Crown Copyright 1995 2

Results Table 1 Percentages of mother tongue Welsh speakers, by assessment of ability For each level of ability, the table shows the proportion whose mother tongue was Welsh and the number in the sample. Percentage of Welsh mother tongue speakers Sample base By assessment of ability: Cannot speak Welsh and never have 0.0 16,897 Cannot speak Welsh but could once 10.3 660 Can only speak a little Welsh 3.6 3,472 Can speak fair amount of Welsh 25.5 877 Fluent in Welsh but never/hardly ever speak it 34.1 160 Fluent in Welsh but speak it only occasionally 58.1 456 Fluent in Welsh and speak it about half the time 63.6 798 Fluent in Welsh and speak it all or most of the time 93.2 2,352 Total current or one time Welsh speakers 33.4 8,775 For entire population 12.6 25,672 Note: percentages based on weighted results, including imputed responses; sample sizes are the number of non-imputed individual responses Table 2 Percentages of mother tongue Welsh speakers, for the whole population, by parental language Percentage of Welsh mother tongue speakers Sample base By parental language: Both parents Welsh speaking 72.8 4,156 Mother only Welsh speaking 13.2 1,491 Father only Welsh speaking 6.0 1,416 Neither parent Welsh speaking 0.4 18,479 People who did not answer questions about the language of their parents 10.9 130 Total 12.6 25,672 Note: percentages based on weighted results, including imputed responses; sample sizes are the number of non-imputed individual responses Table 2 shows that 72.8 per cent of the off-spring of couples where both parents were Welsh speaking were mother tongue speakers. Rates are much lower where only one parent spoke Welsh. This is expected of course as mother tongue speakers come, by definition, from families where Welsh was the foremost language used by the child and this cannot usually be the case if one of the parents does not speak the language. While 72.8 per cent of Welsh speaking couples children are mother tongue speakers, 70.4 per cent of the children claimed to be fluent in 1992 while almost 50 per cent claimed also to speak it all or most of the time (Table 3). (The relevance of analysing families where both parents spoke Welsh is apparent when it is noted that 80.5 per cent of fluent speakers came from such families). Table 3 demonstrates how inter-generational transmission rates vary geographically. Crown Copyright 1995 3

Table 3 Current ability in Welsh of persons with both parents Welsh speaking, by county Per cent Assessment of own ability Clwyd Dyfed Glamorgan: Gwent Gwynedd Powys Wales Mid South West Cannot speak Welsh and never have 10.4 2.3 32.5 16.8 15.8 36.1 0.4 8.1 8.4 Cannot speak Welsh but could once 5.5 1.0 8.3 7.2 5.3 6.9 0.6 3.8 3.1 Can only speak a little Welsh 15.4 7.3 22.9 18.6 15.2 20.9 2.8 12.3 10.2 Can speak fair amount of Welsh 9.1 8.6 11.7 9.3 10.5 11.5 3.1 14.8 7.9 Fluent in Welsh but never/hardly ever speak it or speak it only occasionally 14.7 5.9 11.3 10.5 15.5 14.8 3.8 11.1 8.6 Fluent in Welsh and speak it about half the time 16.1 13.7 4.0 15.7 18.0 3.6 9.4 14.6 12.4 Fluent in Welsh and speak it all or most of the time 28.8 61.2 9.3 21.9 19.7 6.2 79.8 35.3 49.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 99.9 100.0 100.0 Sub-total: at least a little Welsh 84.1 96.7 59.2 76.0 78.9 57.0 98.9 88.1 88.5 Sub-total: fluent 59.6 80.8 24.6 48.1 53.2 24.6 93.0 61.0 70.4 Excludes individuals who did not answer ability question. In interpreting the table it should be borne in mind that the current county of residence is not necessarily where the individual lived as a child and that the transmission rates shown are averages over all ages. The survey results provided some evidence that current transmission rates are higher than in the past. As the table analyses current ability, the rates shown will also mask attrition of high childhood ability and language learning in later life. All Welsh speakers and mother tongue Welsh speakers by age Percentages of mother tongue speakers by age are presented graphically below. Although figures are presented for individual ages, it must be remembered that those individual figures are subject to considerable sampling error: almost all those shown are estimated to have coefficients of variation around 20-25 per cent. The most noticeable features are the high percentages of Welsh speakers among those under 20 years of age and the relatively stable, much lower, percentages of mother tongue speakers in the same age group. Crown Copyright 1995 4

Figure 1 40 Welsh speakers Percentages by age Percentage 30 20 10 0 80 72 64 56 48 40 32 24 16 76 68 60 52 44 36 28 20 12 8 4 All Welsh speakers Mother tongue speakers Age of person The log odds model ln e ((1-y)/y) = 2.6054-0.0155x can be fitted to the proportion of mother tongue speakers, where y is the proportion of mother tongue speakers and x is age. (R Squared = 0.778). It should be noted that this model does not suggest that the proportion of mother tongue speakers will increase in generations yet to be born. The apparent rise in the trend of the percentage of mother tongue speakers amongst those aged under 12 is considered under Validation of results below. Welsh speaking couples Mother tongue speakers of the language come, by definition, from families where Welsh is the foremost language used. The issue of language reproduction within families has been examined from both a sociological and statistical viewpoint (Williams, 1987) and from a psychological viewpoint (Harrison et al., 1981). The statistical examination of Williams was based on the 1981 Census results. He pointed out that language-group exogamy had increased to the point in 1981 that there were almost as many marriages where only one spouse spoke Welsh as there were where both spoke Welsh. By the 1991 Census indeed, amongst families with children aged 3 to 15, there were exactly as many mixed language partnerships as there were Welsh only. Welsh speaking couples by age The social survey permits further examination of this issue of language-group exogamy. The percentage of couples where both partners speak Welsh could be expected to determine the percentage of mother tongue speakers amongst the succeeding generation. The series presented in Table 4 suggest that the percentage of wholly Welsh speaking couples is lower in the younger age groups. This should not be interpreted as showing that younger people are more inclined to marry outside the language community group. Other factors, such as emigration of Welsh speakers and immigration of English speakers, generally greater population movement within Wales, a propensity among Welsh speakers to marry later (a hypothesis only), could all be possible causes of the observed phenomenon. The following is an analysis of all households with both a head of household and a spouse. Of these households containing couples the percentage of them with two Welsh speaking partners was calculated within age groups spanning four years (the age being the average age of the couple). Crown Copyright 1995 5

Table 4 Analysis of all households with both a head of household and a spouse Linguistic mix of couple Number of couples Persons Both Welsh speaking Mixed language couple Both non- Welsh speaking Total Total Welsh speakers as percentage of all Weighted Unweighted sample base persons (a) Average age of couple Per cent Thousands Units Per cent 20-23 6.9 11.4 81.7 100.0 13 156 16.5 24-27 7.3 11.7 81.1 100.0 42 463 14.9 28-31 9.1 10.5 80.4 100.0 57 641 13.8 32-35 9.8 13.0 77.3 100.0 65 723 16.8 36-39 11.0 10.7 78.4 100.0 63 708 16.2 40-43 12.1 11.8 76.1 100.0 64 727 17.5 44-47 10.2 10.5 79.3 100.0 70 791 15.1 48-51 10.1 13.3 76.6 100.0 55 632 18.5 52-55 11.8 12.6 75.6 100.0 51 571 18.2 56-59 12.8 9.9 77.2 100.0 44 502 19.3 60-63 14.6 12.4 73.0 100.0 46 524 20.1 64-67 14.3 12.4 73.3 100.0 45 508 22.1 68-71 14.5 11.9 73.7 100.0 37 424 21.6 72-75 17.1 13.6 69.3 100.0 27 315 24.1 76-80 inclusive 23.4 15.5 61.1 100.0 19 207 27.9 All ages 11.8 12.0 76.2 100.0 708 8,013 20.4 (a) Based on the head of household s assessment Welsh speaking couples - by county and age group The proportions of entirely Welsh speaking couples can be compared to the proportions produced by a simple model which assumes that a partner is selected randomly by an individual from amongst other individuals in the same age group within the same county. The expected proportion of couples where both partners speak Welsh is calculated as the square of the proportion of Welsh speakers in the county age group. The observed and expected proportions are plotted in Figure 2, together with the curves which may be fitted to the points. The fact that the observed proportions are generally higher than the expected proportions under this model seems to indicate the persistence of language-group endogamy. Crown Copyright 1995 6

Figure 2 Actual/expected couples % 70 60 50 40 30 20 10 0 0 Welsh speaking couples Actual and expected percentages 4-year age groups within county 20 40 % of Welsh speakers in county age group 60 80 All Welsh couples Expected proportion Actual proportion Crown Copyright 1995 7

Validation of results The only other official statistics available which distinguish between different types of Welsh speakers relate to primary school pupils. Since 1986 head teachers of primary schools have been asked to make an objective assessment of the number of pupils in each age group: (a) who speak Welsh at home (b) who do not speak Welsh at home but who can speak Welsh with fluency (c) who speak Welsh but not fluently and (d) who cannot speak Welsh at all. Estimates derived from both the primary school statistics and the 1992 Welsh Social Survey are presented in Table 5. The estimates of the percentage of fluent speakers amongst children of primary school age from the social survey and the school statistics are remarkably close, at 14.4 per cent and 14.5 per cent respectively. For the social survey, a design effect was calculated for the proportion of Welsh speakers among the whole population to be 1.6. It seems reasonable to suppose that the design effect for mother tongue speakers in the 5 to 10 age group would be similar. Using this figure gives a 95 per cent confidence interval from the social survey of 6.6 to 10.2 per cent for the estimate of the proportion of fluent mother tongue speakers of Welsh amongst children aged 5 to 10. The figure from Welsh Education Statistics, of 6.8 per cent, lies within this interval. Table 5 Comparison of results for children of primary school age with schools statistics SCHOOLS STATISTICS Assessment of ability Speak Welsh at home Do not speak Welsh at home but who can speak it with fluency PRIMARY SCHOOL PUPILS Number At September, 1992 Per cent of age group: 1992 WELSH SOCIAL SURVEY: Per cent of age group: CHILDREN AGED 5 TO 10 INCLUSIVE Number At September- December 1992 Assessment of ability 14,859 6.8 8.4 19,518 More Welsh than English spoken at home 16,824 7.7 5.9 13,650 Other fluent Total fluent 31,683 14.5 14.4 33,168 Total fluent Speak Welsh but 42,008 19.3 - - not fluently - - 4.1 9,443 a fair amount of Welsh - - 36.8 84,965 a little Welsh Cannot speak Welsh at all 144,401 66.2 44.8 103,532 No Welsh (or didn t answer) Total 218,092 100.0 100.0 231,108 Total 9.3 21,391 Total speaking more Welsh than English at home To make a further comparison with primary school statistics for previous years, estimates can also be derived from the social survey in respect of cohorts spanning 6 years corresponding to primary school Crown Copyright 1995 8

year cohorts. Figure 3 plots these social survey figures of mother tongue speakers with primary school figures of pupils who spoke Welsh at home i. The social survey figures relate to those individuals aged 5 to 10 inclusive in the school year shown. The education statistics are for primary school pupils in the school year. Figure 3 Fluent mother tongue speakers as a percentage of all children aged 5 to 10 incl. 8.5 8.3 8.1 1986 cohort aged 11 to 16 in 1992 7.9 Percentage 7.7 7.5 7.3 7.1 6.9 6.7 1st language pre-1987 Primary school statistics 1992 WSS 6.5 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 1977 School year Each point on the chart represents an estimate of the percentage of all children aged 5 to 10 in the year shown who were fluent mother tongue speakers. The two lines give an indication of how well the assessments of primary school head teachers are reflected by parents (or the individuals own ii ) perceptions of the language ability of their children. One should bear in mind that headteachers may not always be able to judge whether pupils do speak Welsh at home as well as the fact that a non- Welsh speaking parent may not be able to make a proper assessment of their child s ability to speak Welsh. For recent cohorts the correspondence between the two series seems poor with the two coming closer together in respect of children who in 1989 were in primary school. By 1992, the 1989 primary school cohort was aged between 8 and 13. It may be easier for parents to be optimistic about their childrens fluency when their children are younger than 8 and that may be an explanation for the drop from the higher levels recorded for the more recent cohorts by the social survey. The correspondence between the two series is good between 1989 and 1979. The cohort of the 1978 school year was aged 19 to 24 in 1992. There are two reasonable hypotheses as to why the figures derived from the social survey fall from the level of the 1979 cohort. First is population exchange. Large numbers of Welsh children leave Wales at least temporarily when aged around 19, and there is an inflow of students from outside Wales. This population exchange would lower the percentage of mother tongue Welsh speakers amongst 19 to 24 year olds. A second explanation could be that fewer i A change in the question asked of primary schools means that the figures for 1986 and earlier years are for pupils who were described as being fluent first language speakers. ii Individuals aged 16 or over in 1992 answered the questionnaire in their own right: the head of household responded in respect of children under 16. The social survey figures for primary school age cohorts of school years after 1986, aged 15 or under in 1992, are therefore based entirely on the perceptions of the head of households. Figures for cohorts from 1986 to 1982 are based on a mix of individuals and household heads assessments while figures for cohorts before 1982 are based entirely on individuals self assessments. Crown Copyright 1995 9

of the mother tongue speakers classified themselves as fluent by the time they reached this age although a primary school headteacher may have classified them as fluent first language speakers when they were aged 5 to 10 in 1979. In conclusion, despite the differences between the two series, this comparison is generally confirmatory to the results of the social survey. Crown Copyright 1995 10

Technical Annex The derivation of estimates of total numbers of mother tongue speakers The Preliminary Results presented estimates of mother tongue speakers expressed as percentages of all Welsh speakers for both age groups and counties. The estimates of numbers of mother tongue speakers inherent in these percentages were arrived at by simply proportioning the numbers of Welsh speakers who had not answered the question according to the proportion of mother tongue speakers amongst responders. This was done for three language ability categories - little, fair and fluent, within the age group or county concerned. Subsequently, more refined estimates were derived by the use of a sequential hot-deck imputation method (Kalton). This required the specification of classes within which imputation was to occur. In constructing imputation classes the aim was to construct classes of adequate size that explained as much of the variance in the variable to be imputed as possible. The proportion of mother tongue Welsh speakers varied widely between both the ability categories and the parental language mix categories. It was therefore decided that the imputation classes would be defined by two way categorisation according to 7 categories of assessment of own ability and 5 categories of the language mix of the individual's parents. This two way categorisation meant that the spread across the classes of the proportion of mother tongue Welsh speakers was greater than that shown in Tables 1 and 2. Within each class, a single value for the mother language variable subject to imputation was initially assigned at random. The records in the file were then considered in turn. If a record had a response for the variable, its response replaced the value stored for the imputation class in which it fell. If the record had a missing response, it was assigned the value stored for the imputation class. This procedure is similar to randomly assigning values within an imputation class. However, in this survey the records were ordered geographically, by district authority within county, and generally by community within district. There was the potential therefore for the sequential approach of this imputation method to have a useful effect as inter-generational transmission and retention of fluency rates of the language show considerable geographical variation, as shown in Table 3. The 8,775 current or one time Welsh speakers were used to impute values to 1,983 individuals in the 7x 5 classes formed. Individuals who did not speak any Welsh and never had (16,897), and individuals who did not answer the self-categorisation question (65) were treated as separate classes. All the former were imputed not to be mother tongue Welsh speakers, while the latter were counted as mother tongue speakers if they claimed to speak Welsh or if the head of household claimed that they spoke Welsh. The results of the imputation are shown in Table 6. Table 6 Results of imputation Sample base Estimates Initial*: Final : Frequency Per cent Frequency Per cent Frequency Per cent Spoke Welsh at home as a child Yes 3,490 12.6 332,100 12.1 347,600 12.6 No 5,285 19.1 514,500 18.7 696,900 25.3 Not applicable 16,897 61.0 1,706,900 62.0 1,706,900 62.0 Did not answer 2,048 7.4 197,800 7.2 Total 27,720 100.0 2,751,400 100.0 2,751,400 100.0 *After allowing for the survey design and total survey non-response. After imputation For primary school pupils (Table 5), the imputation procedure led to only 8 cases out of 259 nonanswerers being counted as mother tongue speakers. In terms of grossed figures, 732 individuals were imputed as mother tongue speakers. Table 5 reflects the unlikely responses for 20 cases where it was Crown Copyright 1995 11

claimed that although the children spoke more Welsh than English at home though they were not fluent in the language. BIBLIOGRAPHY AITCHISON, J. and CARTER, H. A geography of the Welsh language 1961-1991. Cardiff: University of Wales Press, 1994. HARRISON, G., BELLIN, W. and PIETTE, B. Bilingual mothers in Wales and the language of their children.university of Wales, Board of Celtic Studies, Social Science Monograph No. 6, 1981. JONES, H. A survey of the Welsh language: the 1992 Welsh Social Survey. Statistical News Issue 103, Autumn 1993. HMSO. KALTON,G and KASPRZYK,D. The treatment of missing survey data. Survey Methodology, Vol. 12, No 1, pp 1-16, June 1986. MATHIAS, J. Dwyieithrwydd a r ysgolion Cymraeg. The Welsh Journal of Education Vol.3, No. 2. OPCS. 1991 Census Welsh Language. Wales Topic Monitor (CEN 91 TM WL). 1994. OPCS. 1991 Census Welsh Language Wales (CEN 91 WL). 1994. WELSH OFFICE. 1992 Welsh Social Survey - Preliminary Results, Statistics Press Release, 29 July 1993. WELSH OFFICE. 1992 Welsh Social Survey: Report on the Welsh Language. 1995. WELSH OFFICE. Statistics of Education and Training in Wales: Schools. Last published 1994. WILLIAMS, G. Bilingualism, class dialect and social reproduction. Int l. J. Soc. Lang. 66, 1987. Crown Copyright 1995 12